Context

The US airline carrier 'Falcon Airlines' is facing a decrease in sales over the years, while the airline industry demand is positive growing. therefore, it was imperative that marketing department conducted a survey among 90917 individuals who travelled using the service of the airline, to determine the level of satisfaction based in the service provided, facilities and technology to deliver a better, safe and pleasant experience to the customer.

Hence, the company had established certain parameters which had been considered to play in the important role to understand the consumer demands now a days for better service and be able to identify ways to improve and innovate.

Objective

• Understand which parameters play an important roll in swaying a passanger feedback towards the postive scale.
• Predict wether a passanger will be satisfied or not given the rest of detail are provided.
• Identify which variables contribute for the business profitability and which ones contribute for customer retention

Dictionary

The project has 2 sources of data information, the flight data has information related to the passangers and the performance of the flights in which they travelled and the survey data is the information collected post service experience

Flight data

1. ID: Number of customer
2. Gender: Gender of the passengers (Female, Male).
3. Customer Type: The customer type (Loyal Customer, Disloyal Customer)
4. Age: The actual Age of the passengers
5. Type Travel: Purpose of the passengers'’ flight (Personal or Business travel)
6. Class: travel class in the flight (Business, Economy, Eco Plus)
7. Flight Distance: The flight distance
8. Departure Delay in Mins: Minutes delayed when departure
9. Arrival Delay in Mins: Minutes delayed when arrival

Survey data

1. Satisfaction: Airline satisfaction level (Satisfaction, neutral or dissatisfaction)
2. Seat comfort: Satisfaction level of the seat comfort
3. Departure Arrival Time convenient
4. Food Drink: Satisfaction level of the food and drink
5. Gate Location: Satisfaction level of the gate location
6. Inflight Wi-Fi service: Satisfaction level of the Wi-Fi service.
7. Inflight entertainment: Satisfaction level of the Inflight entertainment
8. Online Support: Satisfaction level of the online support.
9. Ease of Online booking: Satisfaction level of the Ease of Online booking
10. Onboard service: Satisfaction level of the Onboard service
11. Leg room service: Satisfaction level of the leg room service
12. Baggage handling: Satisfaction level of the baggage handling
13. Check in service: Satisfaction level of the checking service
14. Cleanliness: Satisfaction level of the cleanliness
15. Online Boarding: Satisfaction level of the online boarding

Importing Libraries

# this will help in making the Python code more structured automatically (good coding practice)
%load_ext nb_black

# To suppress the warnings
import warnings

warnings.filterwarnings("ignore")


# Libraries to help with reading and manipulating data
import numpy as np
import pandas as pd

# Libraries to help with data visualization
import matplotlib.pyplot as plt

from sklearn import metrics
from sklearn.model_selection import train_test_split, StratifiedKFold, cross_val_score
from sklearn.metrics import (
    f1_score,
    accuracy_score,
    recall_score,
    precision_score,
    confusion_matrix,
    roc_auc_score,
    plot_confusion_matrix,
    plot_roc_curve,
)

%matplotlib inline
import seaborn as sns

# To impute missing values
from sklearn.impute import KNNImputer
from sklearn.impute import SimpleImputer

# For preprocessing
from sklearn import preprocessing

# For Feature selection
from sklearn.feature_selection import SelectKBest, chi2
from sklearn.feature_selection import SelectFromModel
from sklearn.ensemble import RandomForestClassifier as rf
from sklearn.linear_model import LogisticRegression, LogisticRegressionCV, Lasso, lars_path

# For model evaluation
import time

# To help with model building
from sklearn.linear_model import LogisticRegression
from sklearn.tree import DecisionTreeClassifier
from sklearn.ensemble import (
    AdaBoostClassifier,
    GradientBoostingClassifier,
    RandomForestClassifier,
    BaggingClassifier,
)
from xgboost import XGBClassifier

# To suppress scientific notations
pd.set_option("display.float_format", lambda x: "%.3f" % x)

# loading the dataset
df1 = pd.read_csv("Flight data.csv")

# loading the dataset
df2 = pd.read_csv("Surveydata.csv")

df1.shape

(90917, 9)

• Flight data has 90917 rows and 9 columns

df2.shape

(90917, 16)

• Survey data has 90917 rows and 16 columns

# viewing a random sample of the dataset
df1.sample(n=10, random_state=1)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins
50679	200644	Male	Loyal Customer	57	Business travel	Eco	2174	30	36.000
33900	183865	Female	disloyal Customer	35	Business travel	Business	1739	1	7.000
5924	155889	Male	Loyal Customer	56	Personal Travel	Eco	1166	0	0.000
47760	197725	Male	Loyal Customer	58	Business travel	Business	2794	36	60.000
30312	180277	Female	disloyal Customer	21	Business travel	Business	2170	0	0.000
35857	185822	Female	disloyal Customer	25	Business travel	Eco	1868	21	12.000
9051	159016	Male	Loyal Customer	10	Personal Travel	Eco	6787	4	26.000
56968	206933	Female	Loyal Customer	56	Business travel	Eco	446	0	0.000
8993	158958	Female	Loyal Customer	37	Personal Travel	Eco	1960	18	0.000
71828	221793	Male	Loyal Customer	24	Business travel	Business	3761	11	0.000

# viewing a random sample of the dataset
df2.sample(n=10, random_state=1)

	Id	Satisfaction	Seat_comfort	Departure.Arrival.time_convenient	Food_drink	Gate_location	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
50679	200644	satisfied	need improvement	good	good	Convinient	need improvement	need improvement	need improvement	need improvement	acceptable	excellent	good	acceptable	poor	need improvement
33900	183865	neutral or dissatisfied	need improvement	need improvement	NaN	Convinient	poor	need improvement	poor	poor	acceptable	need improvement	good	acceptable	excellent	poor
5924	155889	neutral or dissatisfied	need improvement	poor	extremely poor	Convinient	acceptable	extremely poor	poor	acceptable	need improvement	excellent	good	need improvement	acceptable	acceptable
47760	197725	neutral or dissatisfied	poor	need improvement	need improvement	need improvement	need improvement	good	good	poor	poor	poor	poor	acceptable	poor	poor
30312	180277	neutral or dissatisfied	poor	poor	need improvement	Convinient	need improvement	need improvement	need improvement	need improvement	acceptable	good	acceptable	acceptable	acceptable	need improvement
35857	185822	neutral or dissatisfied	acceptable	poor	acceptable	manageable	excellent	acceptable	excellent	excellent	NaN	excellent	good	acceptable	acceptable	excellent
9051	159016	neutral or dissatisfied	need improvement	good	need improvement	need improvement	need improvement	good	good	excellent	acceptable	good	good	good	good	good
56968	206933	satisfied	acceptable	acceptable	acceptable	manageable	excellent	excellent	poor	acceptable	acceptable	acceptable	acceptable	good	acceptable	poor
8993	158958	satisfied	good	excellent	poor	very convinient	good	poor	excellent	good	good	good	good	good	good	acceptable
71828	221793	satisfied	excellent	excellent	NaN	very convinient	good	good	good	good	good	good	excellent	acceptable	good	good

df2.rename(columns={"Id": "ID"}, inplace=True)

data = df1.merge(df2, on="ID", how="inner")

# viewing a random sample of the dataset
data.sample(n=10, random_state=1)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	...	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
50679	200644	Male	Loyal Customer	57	Business travel	Eco	2174	30	36.000	satisfied	...	need improvement	need improvement	need improvement	need improvement	acceptable	excellent	good	acceptable	poor	need improvement
33900	183865	Female	disloyal Customer	35	Business travel	Business	1739	1	7.000	neutral or dissatisfied	...	poor	need improvement	poor	poor	acceptable	need improvement	good	acceptable	excellent	poor
5924	155889	Male	Loyal Customer	56	Personal Travel	Eco	1166	0	0.000	neutral or dissatisfied	...	acceptable	extremely poor	poor	acceptable	need improvement	excellent	good	need improvement	acceptable	acceptable
47760	197725	Male	Loyal Customer	58	Business travel	Business	2794	36	60.000	neutral or dissatisfied	...	need improvement	good	good	poor	poor	poor	poor	acceptable	poor	poor
30312	180277	Female	disloyal Customer	21	Business travel	Business	2170	0	0.000	neutral or dissatisfied	...	need improvement	need improvement	need improvement	need improvement	acceptable	good	acceptable	acceptable	acceptable	need improvement
35857	185822	Female	disloyal Customer	25	Business travel	Eco	1868	21	12.000	neutral or dissatisfied	...	excellent	acceptable	excellent	excellent	NaN	excellent	good	acceptable	acceptable	excellent
9051	159016	Male	Loyal Customer	10	Personal Travel	Eco	6787	4	26.000	neutral or dissatisfied	...	need improvement	good	good	excellent	acceptable	good	good	good	good	good
56968	206933	Female	Loyal Customer	56	Business travel	Eco	446	0	0.000	satisfied	...	excellent	excellent	poor	acceptable	acceptable	acceptable	acceptable	good	acceptable	poor
8993	158958	Female	Loyal Customer	37	Personal Travel	Eco	1960	18	0.000	satisfied	...	good	poor	excellent	good	good	good	good	good	good	acceptable
71828	221793	Male	Loyal Customer	24	Business travel	Business	3761	11	0.000	satisfied	...	good	good	good	good	good	good	excellent	acceptable	good	good

10 rows × 24 columns

data.shape

(90917, 24)

• The dataset has 90917 rows and 24 columns

# copying the data to another variable to avoid any changes to original data
df = data.copy()

# fixing column names
df.columns = [c.replace(" ", "_") for c in df.columns]

# checking datatypes and number of non-null values for each column
df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 90917 entries, 0 to 90916
Data columns (total 24 columns):
 #   Column                             Non-Null Count  Dtype  
---  ------                             --------------  -----  
 0   ID                                 90917 non-null  int64  
 1   Gender                             90917 non-null  object 
 2   CustomerType                       81818 non-null  object 
 3   Age                                90917 non-null  int64  
 4   TypeTravel                         81829 non-null  object 
 5   Class                              90917 non-null  object 
 6   Flight_Distance                    90917 non-null  int64  
 7   DepartureDelayin_Mins              90917 non-null  int64  
 8   ArrivalDelayin_Mins                90633 non-null  float64
 9   Satisfaction                       90917 non-null  object 
 10  Seat_comfort                       90917 non-null  object 
 11  Departure.Arrival.time_convenient  82673 non-null  object 
 12  Food_drink                         82736 non-null  object 
 13  Gate_location                      90917 non-null  object 
 14  Inflightwifi_service               90917 non-null  object 
 15  Inflight_entertainment             90917 non-null  object 
 16  Online_support                     90917 non-null  object 
 17  Ease_of_Onlinebooking              90917 non-null  object 
 18  Onboard_service                    83738 non-null  object 
 19  Leg_room_service                   90917 non-null  object 
 20  Baggage_handling                   90917 non-null  object 
 21  Checkin_service                    90917 non-null  object 
 22  Cleanliness                        90917 non-null  object 
 23  Online_boarding                    90917 non-null  object 
dtypes: float64(1), int64(4), object(19)
memory usage: 17.3+ MB

• Dataset has 90917 survey entries and passanger flight details from Falcon Airlines. In total are 22 feature columns and 1 binary target column .
• ID, Age, Flight distance, Departure Delay in Mins and Arrival Delay in Mins are numerical variables, the others are decribed as object.
• Customer Type, Type Travel, Arrival Delay in Mins, Departure arrivel time convenient, Food drink and Onboard service paramenters have missing values.

# let's check for duplicate values in the data
df.duplicated().sum()

0

• No Duplicates in the dataset

# let's check for missing values in the data
pd.DataFrame(
    data={"% of Missing Values": round(df.isna().sum() / df.isna().count() * 100, 2)}
)

	% of Missing Values
ID	0.000
Gender	0.000
CustomerType	10.010
Age	0.000
TypeTravel	10.000
Class	0.000
Flight_Distance	0.000
DepartureDelayin_Mins	0.000
ArrivalDelayin_Mins	0.310
Satisfaction	0.000
Seat_comfort	0.000
Departure.Arrival.time_convenient	9.070
Food_drink	9.000
Gate_location	0.000
Inflightwifi_service	0.000
Inflight_entertainment	0.000
Online_support	0.000
Ease_of_Onlinebooking	0.000
Onboard_service	7.900
Leg_room_service	0.000
Baggage_handling	0.000
Checkin_service	0.000
Cleanliness	0.000
Online_boarding	0.000

• Customer type 10.01% of missing values
• Type Travel 10.00% of missing values
• Arrival Delay in Mins 0.31% of missing values
• Departure arrivel time convenient 9.07% of missing values
• Food drink 9.00% of missing values
• Onboard service 7.90% of missing values

# Checking missing values per row
num_missing = df.isnull().sum(axis=1)
num_missing.value_counts()

0    53634
1    32503
2     4768
3       12
dtype: int64

df[num_missing == 3].sample(n=10)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	...	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
31790	181755	Female	NaN	24	Business travel	Eco	1844	23	NaN	neutral or dissatisfied	...	need improvement	need improvement	acceptable	need improvement	NaN	poor	need improvement	acceptable	acceptable	need improvement
72591	222556	Female	NaN	37	Business travel	Business	4019	0	NaN	neutral or dissatisfied	...	need improvement	good	good	good	NaN	good	good	poor	good	good
87450	237415	Female	Loyal Customer	50	NaN	Business	1753	19	NaN	satisfied	...	excellent	good	excellent	excellent	NaN	excellent	excellent	good	excellent	excellent
85743	235708	Male	Loyal Customer	50	NaN	Business	676	4	NaN	satisfied	...	need improvement	excellent	good	excellent	excellent	excellent	excellent	excellent	excellent	acceptable
56851	206816	Male	Loyal Customer	58	NaN	Eco	2009	55	NaN	neutral or dissatisfied	...	acceptable	acceptable	acceptable	acceptable	acceptable	poor	need improvement	good	need improvement	acceptable
1629	151594	Male	NaN	8	Personal Travel	Eco	3632	0	NaN	neutral or dissatisfied	...	poor	extremely poor	poor	poor	NaN	good	acceptable	poor	acceptable	poor
50518	200483	Male	NaN	60	Business travel	Eco Plus	1333	13	NaN	neutral or dissatisfied	...	need improvement	need improvement	need improvement	need improvement	acceptable	excellent	need improvement	acceptable	acceptable	need improvement
59655	209620	Male	Loyal Customer	42	NaN	Eco	1526	73	NaN	neutral or dissatisfied	...	acceptable	acceptable	acceptable	acceptable	poor	need improvement	acceptable	poor	acceptable	acceptable
28843	178808	Female	disloyal Customer	24	NaN	Eco Plus	1798	0	NaN	neutral or dissatisfied	...	good	poor	good	good	NaN	poor	good	acceptable	good	good
72277	222242	Female	Loyal Customer	25	NaN	Business	2736	21	NaN	satisfied	...	good	good	good	good	excellent	excellent	excellent	acceptable	excellent	good

10 rows × 24 columns

df[num_missing == 2].sample(n=10)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	...	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
15052	165017	Male	NaN	11	Personal Travel	Eco	2818	100	92.000	neutral or dissatisfied	...	poor	acceptable	poor	poor	need improvement	need improvement	poor	good	need improvement	poor
6660	156625	Male	Loyal Customer	53	NaN	Eco	2006	22	54.000	neutral or dissatisfied	...	poor	poor	poor	poor	NaN	acceptable	acceptable	poor	need improvement	poor
35964	185929	Male	disloyal Customer	36	NaN	Eco	1955	22	19.000	neutral or dissatisfied	...	acceptable	acceptable	acceptable	acceptable	NaN	poor	acceptable	need improvement	acceptable	acceptable
87496	237461	Female	NaN	53	Business travel	Business	1624	3	8.000	satisfied	...	good	good	good	excellent	NaN	excellent	excellent	acceptable	excellent	acceptable
48902	198867	Male	Loyal Customer	56	NaN	Business	1951	1	0.000	neutral or dissatisfied	...	need improvement	good	good	poor	poor	need improvement	poor	need improvement	poor	acceptable
21963	171928	Male	NaN	60	Personal Travel	Eco	1630	7	0.000	neutral or dissatisfied	...	poor	acceptable	acceptable	poor	excellent	need improvement	need improvement	good	acceptable	poor
42612	192577	Female	NaN	23	Business travel	Eco	1887	20	15.000	satisfied	...	excellent	excellent	excellent	excellent	NaN	excellent	excellent	acceptable	good	excellent
23474	173439	Male	Loyal Customer	33	NaN	Eco	4354	0	0.000	satisfied	...	good	need improvement	need improvement	acceptable	excellent	need improvement	need improvement	need improvement	acceptable	need improvement
12444	162409	Male	NaN	12	Personal Travel	Eco	2585	6	0.000	neutral or dissatisfied	...	good	need improvement	good	good	NaN	acceptable	excellent	good	excellent	good
43551	193516	Female	disloyal Customer	25	NaN	Business	1792	0	0.000	satisfied	...	poor	excellent	poor	poor	NaN	good	good	excellent	excellent	poor

10 rows × 24 columns

df[num_missing == 1].sample(n=10)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	...	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
33861	183826	Female	disloyal Customer	36	Business travel	Business	2105	32	7.000	neutral or dissatisfied	...	need improvement	need improvement	need improvement	need improvement	good	good	good	acceptable	good	need improvement
75342	225307	Female	Loyal Customer	60	Business travel	Business	921	5	14.000	satisfied	...	good	excellent	excellent	good	NaN	good	good	excellent	good	excellent
46791	196756	Female	NaN	58	Business travel	Business	1977	0	0.000	neutral or dissatisfied	...	good	good	acceptable	poor	poor	poor	poor	need improvement	poor	need improvement
82093	232058	Male	Loyal Customer	42	Business travel	Business	1920	36	22.000	satisfied	...	need improvement	acceptable	good	excellent	excellent	excellent	excellent	good	excellent	excellent
50036	200001	Male	Loyal Customer	43	Business travel	Eco	1991	0	6.000	neutral or dissatisfied	...	need improvement	need improvement	need improvement	need improvement	NaN	good	poor	acceptable	need improvement	need improvement
89190	239155	Male	Loyal Customer	52	Business travel	Business	353	10	13.000	satisfied	...	good	good	good	excellent	excellent	excellent	excellent	excellent	excellent	acceptable
60690	210655	Female	Loyal Customer	50	Business travel	Business	1359	0	0.000	neutral or dissatisfied	...	excellent	acceptable	good	good	NaN	acceptable	good	acceptable	good	need improvement
48582	198547	Male	Loyal Customer	18	NaN	Eco	2035	26	39.000	neutral or dissatisfied	...	poor	poor	acceptable	poor	acceptable	acceptable	good	need improvement	acceptable	poor
76006	225971	Male	Loyal Customer	42	Business travel	Business	406	0	5.000	satisfied	...	acceptable	excellent	good	good	good	good	good	good	good	excellent
90742	240707	Female	NaN	58	Business travel	Business	3857	0	0.000	satisfied	...	excellent	good	good	excellent	excellent	excellent	excellent	good	excellent	excellent

10 rows × 24 columns

# checking the number of unique values in each column
df.nunique()

ID                                   90917
Gender                                   2
CustomerType                             2
Age                                     75
TypeTravel                               2
Class                                    3
Flight_Distance                       5213
DepartureDelayin_Mins                  436
ArrivalDelayin_Mins                    445
Satisfaction                             2
Seat_comfort                             6
Departure.Arrival.time_convenient        6
Food_drink                               6
Gate_location                            6
Inflightwifi_service                     6
Inflight_entertainment                   6
Online_support                           6
Ease_of_Onlinebooking                    6
Onboard_service                          6
Leg_room_service                         6
Baggage_handling                         5
Checkin_service                          6
Cleanliness                              6
Online_boarding                          6
dtype: int64

• Baggage handling has 1 level of satisfaction less than the other parameters.

df.describe().T

	count	mean	std	min	25%	50%	75%	max
ID	90917.000	195423.000	26245.622	149965.000	172694.000	195423.000	218152.000	240881.000
Age	90917.000	39.447	15.130	7.000	27.000	40.000	51.000	85.000
Flight_Distance	90917.000	1981.629	1026.780	50.000	1360.000	1927.000	2542.000	6950.000
DepartureDelayin_Mins	90917.000	14.687	38.669	0.000	0.000	0.000	12.000	1592.000
ArrivalDelayin_Mins	90633.000	15.059	39.039	0.000	0.000	0.000	13.000	1584.000

• Age: does from 7 to 85 years old. Mean and median is around 40, suggesting a normal distribution.
• Flight Distance from 50 to 6950. From mean and median is suggesting a normal distribution.
• Departure Delay in Mins: most of the data suggest not to have delays, however it seems to have extreme outliers.
• Arrival Dlay in Mins: most of the data suggest not to have delays, however it seems to have extreme outliers.

df.describe(include=["object"]).T

	count	unique	top	freq
Gender	90917	2	Female	46186
CustomerType	81818	2	Loyal Customer	66897
TypeTravel	81829	2	Business travel	56481
Class	90917	3	Business	43535
Satisfaction	90917	2	satisfied	49761
Seat_comfort	90917	6	acceptable	20552
Departure.Arrival.time_convenient	82673	6	good	18840
Food_drink	82736	6	acceptable	17991
Gate_location	90917	6	manageable	23385
Inflightwifi_service	90917	6	good	22159
Inflight_entertainment	90917	6	good	29373
Online_support	90917	6	good	29042
Ease_of_Onlinebooking	90917	6	good	27993
Onboard_service	83738	6	good	26373
Leg_room_service	90917	6	good	27814
Baggage_handling	90917	5	good	33822
Checkin_service	90917	6	good	25483
Cleanliness	90917	6	good	34246
Online_boarding	90917	6	good	24676

• Gender: Most of the customers are female
• Customer Type Most of the customers are loyals
• Type Travel: most of the customers has a bussiness purpose
• Class: most of the customer where in business class
• Satisfaction: most of the travellers are satisfied with the service.
• Seat comfort and Food drink are in satisfection level acceptable for most of the customers.
• Departure arrival time, Inflight Wi-Fi service, Inflight entertainment, Leg room service, Cleanliness, Gate Location, Onboard service, Baggage handling, Check in service are in satisfaction level good for most of the customers.
• Gate location: most of the customers say that it is in manageable level.

for i in df.describe(include=["object"]).columns:
    print("Unique values in", i, "are :")
    print(df[i].value_counts())
    print("*" * 50)

Unique values in Gender are :
Female    46186
Male      44731
Name: Gender, dtype: int64
**************************************************
Unique values in CustomerType are :
Loyal Customer       66897
disloyal Customer    14921
Name: CustomerType, dtype: int64
**************************************************
Unique values in TypeTravel are :
Business travel    56481
Personal Travel    25348
Name: TypeTravel, dtype: int64
**************************************************
Unique values in Class are :
Business    43535
Eco         40758
Eco Plus     6624
Name: Class, dtype: int64
**************************************************
Unique values in Satisfaction are :
satisfied                  49761
neutral or dissatisfied    41156
Name: Satisfaction, dtype: int64
**************************************************
Unique values in Seat_comfort are :
acceptable          20552
need improvement    20002
good                19789
poor                14687
excellent           12519
extremely poor       3368
Name: Seat_comfort, dtype: int64
**************************************************
Unique values in Departure.Arrival.time_convenient are :
good                18840
excellent           17079
acceptable          14806
need improvement    14539
poor                13210
extremely poor       4199
Name: Departure.Arrival.time_convenient, dtype: int64
**************************************************
Unique values in Food_drink are :
acceptable          17991
need improvement    17359
good                17245
poor                13400
excellent           12947
extremely poor       3794
Name: Food_drink, dtype: int64
**************************************************
Unique values in Gate_location are :
manageable           23385
Convinient           21088
need improvement     17113
Inconvinient         15876
very convinient      13454
very inconvinient        1
Name: Gate_location, dtype: int64
**************************************************
Unique values in Inflightwifi_service are :
good                22159
excellent           20258
acceptable          19199
need improvement    18894
poor                10311
extremely poor         96
Name: Inflightwifi_service, dtype: int64
**************************************************
Unique values in Inflight_entertainment are :
good                29373
excellent           20786
acceptable          16995
need improvement    13527
poor                 8198
extremely poor       2038
Name: Inflight_entertainment, dtype: int64
**************************************************
Unique values in Online_support are :
good                29042
excellent           24916
acceptable          15090
need improvement    12063
poor                 9805
extremely poor          1
Name: Online_support, dtype: int64
**************************************************
Unique values in Ease_of_Onlinebooking are :
good                27993
excellent           23960
acceptable          15686
need improvement    13896
poor                 9370
extremely poor         12
Name: Ease_of_Onlinebooking, dtype: int64
**************************************************
Unique values in Onboard_service are :
good                26373
excellent           20396
acceptable          17411
need improvement    11018
poor                 8537
extremely poor          3
Name: Onboard_service, dtype: int64
**************************************************
Unique values in Leg_room_service are :
good                27814
excellent           24071
acceptable          15775
need improvement    15156
poor                 7779
extremely poor        322
Name: Leg_room_service, dtype: int64
**************************************************
Unique values in Baggage_handling are :
good                33822
excellent           25002
acceptable          17233
need improvement     9301
poor                 5559
Name: Baggage_handling, dtype: int64
**************************************************
Unique values in Checkin_service are :
good                25483
acceptable          24941
excellent           18918
need improvement    10813
poor                10761
extremely poor          1
Name: Checkin_service, dtype: int64
**************************************************
Unique values in Cleanliness are :
good                34246
excellent           25079
acceptable          16930
need improvement     9283
poor                 5375
extremely poor          4
Name: Cleanliness, dtype: int64
**************************************************
Unique values in Online_boarding are :
good                24676
acceptable          21427
excellent           20993
need improvement    13035
poor                10777
extremely poor          9
Name: Online_boarding, dtype: int64
**************************************************

Data Pre-processing

The survey parameteres and binary target needs to be encode for easier syntax.

## Encoding the survay data

df["Satisfaction"].replace("neutral or dissatisfied", 0, inplace=True)
df["Satisfaction"].replace("satisfied", 1, inplace=True)

df["Seat_comfort"].replace("extremely poor", 0, inplace=True)
df["Seat_comfort"].replace("poor", 1, inplace=True)
df["Seat_comfort"].replace("need improvement", 2, inplace=True)
df["Seat_comfort"].replace("acceptable", 3, inplace=True)
df["Seat_comfort"].replace("good", 4, inplace=True)
df["Seat_comfort"].replace("excellent", 5, inplace=True)

df["Departure.Arrival.time_convenient"].replace("extremely poor", 0, inplace=True)
df["Departure.Arrival.time_convenient"].replace("poor", 1, inplace=True)
df["Departure.Arrival.time_convenient"].replace("need improvement", 2, inplace=True)
df["Departure.Arrival.time_convenient"].replace("acceptable", 3, inplace=True)
df["Departure.Arrival.time_convenient"].replace("good", 4, inplace=True)
df["Departure.Arrival.time_convenient"].replace("excellent", 5, inplace=True)

df["Food_drink"].replace("extremely poor", 0, inplace=True)
df["Food_drink"].replace("poor", 1, inplace=True)
df["Food_drink"].replace("need improvement", 2, inplace=True)
df["Food_drink"].replace("acceptable", 3, inplace=True)
df["Food_drink"].replace("good", 4, inplace=True)
df["Food_drink"].replace("excellent", 5, inplace=True)

df["Gate_location"].replace("very inconvinient", 0, inplace=True)
df["Gate_location"].replace("Inconvinient", 1, inplace=True)
df["Gate_location"].replace("need improvement", 2, inplace=True)
df["Gate_location"].replace("manageable", 3, inplace=True)
df["Gate_location"].replace("Convinient", 4, inplace=True)
df["Gate_location"].replace("very convinient", 5, inplace=True)

df["Inflightwifi_service"].replace("extremely poor", 0, inplace=True)
df["Inflightwifi_service"].replace("poor", 1, inplace=True)
df["Inflightwifi_service"].replace("need improvement", 2, inplace=True)
df["Inflightwifi_service"].replace("acceptable", 3, inplace=True)
df["Inflightwifi_service"].replace("good", 4, inplace=True)
df["Inflightwifi_service"].replace("excellent", 5, inplace=True)

df["Inflight_entertainment"].replace("extremely poor", 0, inplace=True)
df["Inflight_entertainment"].replace("poor", 1, inplace=True)
df["Inflight_entertainment"].replace("need improvement", 2, inplace=True)
df["Inflight_entertainment"].replace("acceptable", 3, inplace=True)
df["Inflight_entertainment"].replace("good", 4, inplace=True)
df["Inflight_entertainment"].replace("excellent", 5, inplace=True)

df["Online_support"].replace("extremely poor", 0, inplace=True)
df["Online_support"].replace("poor", 1, inplace=True)
df["Online_support"].replace("need improvement", 2, inplace=True)
df["Online_support"].replace("acceptable", 3, inplace=True)
df["Online_support"].replace("good", 4, inplace=True)
df["Online_support"].replace("excellent", 5, inplace=True)

df["Ease_of_Onlinebooking"].replace("extremely poor", 0, inplace=True)
df["Ease_of_Onlinebooking"].replace("poor", 1, inplace=True)
df["Ease_of_Onlinebooking"].replace("need improvement", 2, inplace=True)
df["Ease_of_Onlinebooking"].replace("acceptable", 3, inplace=True)
df["Ease_of_Onlinebooking"].replace("good", 4, inplace=True)
df["Ease_of_Onlinebooking"].replace("excellent", 5, inplace=True)

df["Onboard_service"].replace("extremely poor", 0, inplace=True)
df["Onboard_service"].replace("poor", 1, inplace=True)
df["Onboard_service"].replace("need improvement", 2, inplace=True)
df["Onboard_service"].replace("acceptable", 3, inplace=True)
df["Onboard_service"].replace("good", 4, inplace=True)
df["Onboard_service"].replace("excellent", 5, inplace=True)


df["Leg_room_service"].replace("extremely poor", 0, inplace=True)
df["Leg_room_service"].replace("poor", 1, inplace=True)
df["Leg_room_service"].replace("need improvement", 2, inplace=True)
df["Leg_room_service"].replace("acceptable", 3, inplace=True)
df["Leg_room_service"].replace("good", 4, inplace=True)
df["Leg_room_service"].replace("excellent", 5, inplace=True)

df["Checkin_service"].replace("extremely poor", 0, inplace=True)
df["Checkin_service"].replace("poor", 1, inplace=True)
df["Checkin_service"].replace("need improvement", 2, inplace=True)
df["Checkin_service"].replace("acceptable", 3, inplace=True)
df["Checkin_service"].replace("good", 4, inplace=True)
df["Checkin_service"].replace("excellent", 5, inplace=True)

df["Cleanliness"].replace("extremely poor", 0, inplace=True)
df["Cleanliness"].replace("poor", 1, inplace=True)
df["Cleanliness"].replace("need improvement", 2, inplace=True)
df["Cleanliness"].replace("acceptable", 3, inplace=True)
df["Cleanliness"].replace("good", 4, inplace=True)
df["Cleanliness"].replace("excellent", 5, inplace=True)

df["Online_boarding"].replace("extremely poor", 0, inplace=True)
df["Online_boarding"].replace("poor", 1, inplace=True)
df["Online_boarding"].replace("need improvement", 2, inplace=True)
df["Online_boarding"].replace("acceptable", 3, inplace=True)
df["Online_boarding"].replace("good", 4, inplace=True)
df["Online_boarding"].replace("excellent", 5, inplace=True)


df["Baggage_handling"].replace("poor", 1, inplace=True)
df["Baggage_handling"].replace("need improvement", 2, inplace=True)
df["Baggage_handling"].replace("acceptable", 3, inplace=True)
df["Baggage_handling"].replace("good", 4, inplace=True)
df["Baggage_handling"].replace("excellent", 5, inplace=True)

# viewing a random sample of the dataset
df.sample(n=10, random_state=1)

	ID	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	...	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding
50679	200644	Male	Loyal Customer	57	Business travel	Eco	2174	30	36.000	1	...	2	2	2	2	3.000	5	4	3	1	2
33900	183865	Female	disloyal Customer	35	Business travel	Business	1739	1	7.000	0	...	1	2	1	1	3.000	2	4	3	5	1
5924	155889	Male	Loyal Customer	56	Personal Travel	Eco	1166	0	0.000	0	...	3	0	1	3	2.000	5	4	2	3	3
47760	197725	Male	Loyal Customer	58	Business travel	Business	2794	36	60.000	0	...	2	4	4	1	1.000	1	1	3	1	1
30312	180277	Female	disloyal Customer	21	Business travel	Business	2170	0	0.000	0	...	2	2	2	2	3.000	4	3	3	3	2
35857	185822	Female	disloyal Customer	25	Business travel	Eco	1868	21	12.000	0	...	5	3	5	5	NaN	5	4	3	3	5
9051	159016	Male	Loyal Customer	10	Personal Travel	Eco	6787	4	26.000	0	...	2	4	4	5	3.000	4	4	4	4	4
56968	206933	Female	Loyal Customer	56	Business travel	Eco	446	0	0.000	1	...	5	5	1	3	3.000	3	3	4	3	1
8993	158958	Female	Loyal Customer	37	Personal Travel	Eco	1960	18	0.000	1	...	4	1	5	4	4.000	4	4	4	4	3
71828	221793	Male	Loyal Customer	24	Business travel	Business	3761	11	0.000	1	...	4	4	4	4	4.000	4	5	3	4	4

10 rows × 24 columns

# ID consists of uniques ID for clients and hence will not add value to the modeling
df.drop(["ID"], axis=1, inplace=True)

# Making a list of all catrgorical variables
cat_col = [
    "Gender",
    "CustomerType",
    "TypeTravel",
    "Class",
]
# Converting the data type of each categorical variable to 'category'
for column in cat_col:
    df[column] = df[column].astype("category")

df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 90917 entries, 0 to 90916
Data columns (total 23 columns):
 #   Column                             Non-Null Count  Dtype   
---  ------                             --------------  -----   
 0   Gender                             90917 non-null  category
 1   CustomerType                       81818 non-null  category
 2   Age                                90917 non-null  int64   
 3   TypeTravel                         81829 non-null  category
 4   Class                              90917 non-null  category
 5   Flight_Distance                    90917 non-null  int64   
 6   DepartureDelayin_Mins              90917 non-null  int64   
 7   ArrivalDelayin_Mins                90633 non-null  float64 
 8   Satisfaction                       90917 non-null  int64   
 9   Seat_comfort                       90917 non-null  int64   
 10  Departure.Arrival.time_convenient  82673 non-null  float64 
 11  Food_drink                         82736 non-null  float64 
 12  Gate_location                      90917 non-null  int64   
 13  Inflightwifi_service               90917 non-null  int64   
 14  Inflight_entertainment             90917 non-null  int64   
 15  Online_support                     90917 non-null  int64   
 16  Ease_of_Onlinebooking              90917 non-null  int64   
 17  Onboard_service                    83738 non-null  float64 
 18  Leg_room_service                   90917 non-null  int64   
 19  Baggage_handling                   90917 non-null  int64   
 20  Checkin_service                    90917 non-null  int64   
 21  Cleanliness                        90917 non-null  int64   
 22  Online_boarding                    90917 non-null  int64   
dtypes: category(4), float64(4), int64(15)
memory usage: 14.2 MB

Columns Processing

The data columns will be splitted according to identify variables that are monitored by the company, and which are controlled by the business. so the parameters will be splitted in 2.

Variables that wants to be controlled to retain customers:

• Customers Information

  • ID
    • Gender
    • Customer Type
    • Age
    • Type Travel
    • Class
    • Flight Distance
    • Departure Delay in Mins
    • Arrival Delay in Mins

And Variables controlled by the business which have an impact in to the company profits:

• Facilities Satisfaction

  • Gate Location
  • Onboard service
  • Baggage handling
  • Check in service

• Online service Satisfaction

  • Online Support
  • Ease of Online booking
  • Online Boarding

• In Flight Satisfaction

  • Seat comfort
  • Departure Arrival Time convenient
  • Food Drink
  • Inflight Wi-Fi service
  • Inflight entertainment
  • Leg room service
  • Cleanliness

  These features metioned above belong to the survay entries where passangers rate their flight experience on a sclae 0 to 5

### Create a new column based on the facility average
df["Facilities_avg"]= df[["Gate_location", "Onboard_service", "Baggage_handling", "Checkin_service"]].mean(axis=1)

### Create a new column based on the online service average
df["Online_service_avg"] = df[
    ["Online_support", "Ease_of_Onlinebooking", "Online_boarding"]
].mean(axis=1)

### Create a new column based on the Inflight average
df["InFlight_avg"] = df[
    [
        "Seat_comfort",
        "Departure.Arrival.time_convenient",
        "Food_drink",
        "Inflightwifi_service",
        "Inflight_entertainment",
        "Leg_room_service",
        "Cleanliness",
    ]
].mean(axis=1)

# round the aveage results

df.Facilities_avg = df.Facilities_avg.round()
df.Online_service_avg = df.Online_service_avg.round()
df.InFlight_avg = df.InFlight_avg.round()

# viewing a random sample of the sumary dataset
df[
    [
        "Gender",
        "CustomerType",
        "Age",
        "TypeTravel",
        "Class",
        "Flight_Distance",
        "DepartureDelayin_Mins",
        "ArrivalDelayin_Mins",
        "Satisfaction",
        "Facilities_avg",
        "Online_service_avg",
        "InFlight_avg",
    ]
].sample(n=10, random_state=1)

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Facilities_avg	Online_service_avg	InFlight_avg
50679	Male	Loyal Customer	57	Business travel	Eco	2174	30	36.000	1	4.000	2.000	3.000
33900	Female	disloyal Customer	35	Business travel	Business	1739	1	7.000	0	4.000	1.000	2.000
5924	Male	Loyal Customer	56	Personal Travel	Eco	1166	0	0.000	0	3.000	2.000	2.000
47760	Male	Loyal Customer	58	Business travel	Business	2794	36	60.000	0	2.000	2.000	2.000
30312	Female	disloyal Customer	21	Business travel	Business	2170	0	0.000	0	3.000	2.000	2.000
35857	Female	disloyal Customer	25	Business travel	Eco	1868	21	12.000	0	3.000	5.000	3.000
9051	Male	Loyal Customer	10	Personal Travel	Eco	6787	4	26.000	0	3.000	4.000	3.000
56968	Female	Loyal Customer	56	Business travel	Eco	446	0	0.000	1	3.000	2.000	4.000
8993	Female	Loyal Customer	37	Personal Travel	Eco	1960	18	0.000	1	4.000	4.000	3.000
71828	Male	Loyal Customer	24	Business travel	Business	3761	11	0.000	1	4.000	4.000	4.000

df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 90917 entries, 0 to 90916
Data columns (total 26 columns):
 #   Column                             Non-Null Count  Dtype   
---  ------                             --------------  -----   
 0   Gender                             90917 non-null  category
 1   CustomerType                       81818 non-null  category
 2   Age                                90917 non-null  int64   
 3   TypeTravel                         81829 non-null  category
 4   Class                              90917 non-null  category
 5   Flight_Distance                    90917 non-null  int64   
 6   DepartureDelayin_Mins              90917 non-null  int64   
 7   ArrivalDelayin_Mins                90633 non-null  float64 
 8   Satisfaction                       90917 non-null  int64   
 9   Seat_comfort                       90917 non-null  int64   
 10  Departure.Arrival.time_convenient  82673 non-null  float64 
 11  Food_drink                         82736 non-null  float64 
 12  Gate_location                      90917 non-null  int64   
 13  Inflightwifi_service               90917 non-null  int64   
 14  Inflight_entertainment             90917 non-null  int64   
 15  Online_support                     90917 non-null  int64   
 16  Ease_of_Onlinebooking              90917 non-null  int64   
 17  Onboard_service                    83738 non-null  float64 
 18  Leg_room_service                   90917 non-null  int64   
 19  Baggage_handling                   90917 non-null  int64   
 20  Checkin_service                    90917 non-null  int64   
 21  Cleanliness                        90917 non-null  int64   
 22  Online_boarding                    90917 non-null  int64   
 23  Facilities_avg                     90917 non-null  float64 
 24  Online_service_avg                 90917 non-null  float64 
 25  InFlight_avg                       90917 non-null  float64 
dtypes: category(4), float64(7), int64(15)
memory usage: 16.3 MB

• New variables added to the data set

Facilities Average Stadistics Summary

df[
    [
        "Facilities_avg",
        "Gate_location",
        "Onboard_service",
        "Baggage_handling",
        "Checkin_service",
        "Satisfaction",
    ]
].describe()

	Facilities_avg	Gate_location	Onboard_service	Baggage_handling	Checkin_service	Satisfaction
count	90917.000	90917.000	83738.000	90917.000	90917.000	90917.000
mean	3.376	2.990	3.467	3.697	3.341	0.547
std	0.816	1.308	1.269	1.154	1.261	0.498
min	1.000	0.000	0.000	1.000	0.000	0.000
25%	3.000	2.000	3.000	3.000	3.000	0.000
50%	4.000	3.000	4.000	4.000	3.000	1.000
75%	4.000	4.000	4.000	5.000	4.000	1.000
max	5.000	5.000	5.000	5.000	5.000	1.000

• Overall, most of the people think the Facilities are good.

Online Service Average Stadistics Summary

df[
    [
        "Online_service_avg",
        "Online_support",
        "Ease_of_Onlinebooking",
        "Online_boarding",
        "Satisfaction",
    ]
].describe()

	Online_service_avg	Online_support	Ease_of_Onlinebooking	Online_boarding	Satisfaction
count	90917.000	90917.000	90917.000	90917.000	90917.000
mean	3.449	3.519	3.476	3.352	0.547
std	1.165	1.308	1.305	1.300	0.498
min	1.000	0.000	0.000	0.000	0.000
25%	3.000	3.000	2.000	2.000	0.000
50%	4.000	4.000	4.000	4.000	1.000
75%	4.000	5.000	5.000	4.000	1.000
max	5.000	5.000	5.000	5.000	1.000

• Overall, most of the people think the online services are good.

InFlight Average Stadistics Summary

df[
    [
        "InFlight_avg",
        "Seat_comfort",
        "Departure.Arrival.time_convenient",
        "Food_drink",
        "Inflightwifi_service",
        "Inflight_entertainment",
        "Leg_room_service",
        "Cleanliness",
        "Satisfaction",
    ]
].describe()

	InFlight_avg	Seat_comfort	Departure.Arrival.time_convenient	Food_drink	Inflightwifi_service	Inflight_entertainment	Leg_room_service	Cleanliness	Satisfaction
count	90917.000	90917.000	82673.000	82736.000	90917.000	90917.000	90917.000	90917.000	90917.000
mean	3.227	2.839	2.993	2.850	3.252	3.384	3.487	3.708	0.547
std	0.816	1.394	1.525	1.443	1.320	1.342	1.292	1.148	0.498
min	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000	0.000
25%	3.000	2.000	2.000	2.000	2.000	2.000	2.000	3.000	0.000
50%	3.000	3.000	3.000	3.000	3.000	4.000	4.000	4.000	1.000
75%	4.000	4.000	4.000	4.000	4.000	4.000	5.000	5.000	1.000
max	5.000	5.000	5.000	5.000	5.000	5.000	5.000	5.000	1.000

• Overall, most of the people think the Inflight service is acceptable.

EDA

Univariate Analysis

# function to plot a boxplot and a histogram along the same scale.


def histogram_boxplot(data, feature, figsize=(12, 7), kde=False, bins=None):
    """
    Boxplot and histogram combined

    data: dataframe
    feature: dataframe column
    figsize: size of figure (default (12,7))
    kde: whether to the show density curve (default False)
    bins: number of bins for histogram (default None)
    """
    f2, (ax_box2, ax_hist2) = plt.subplots(
        nrows=2,  # Number of rows of the subplot grid= 2
        sharex=True,  # x-axis will be shared among all subplots
        gridspec_kw={"height_ratios": (0.25, 0.75)},
        figsize=figsize,
    )  # creating the 2 subplots
    sns.boxplot(
        data=data, x=feature, ax=ax_box2, showmeans=True, color="violet"
    )  # boxplot will be created and a star will indicate the mean value of the column
    sns.histplot(
        data=data, x=feature, kde=kde, ax=ax_hist2, bins=bins, palette="winter"
    ) if bins else sns.histplot(
        data=data, x=feature, kde=kde, ax=ax_hist2
    )  # For histogram
    ax_hist2.axvline(
        data[feature].mean(), color="green", linestyle="--"
    )  # Add mean to the histogram
    ax_hist2.axvline(
        data[feature].median(), color="black", linestyle="-"
    )  # Add median to the histogram

Observations from Age

histogram_boxplot(df, "Age")

• the plot suggest a normal distribution.
• there are to 2 peaks around ~27 and ~46 years old, 50% of the polulation are between 27 and 50 yars old.
• It is imperative to check the kids filling out the survey under 15, probably it is an error or parents filling the survey for their kids.
• From boxplot not outliers showed.

df[df["Age"] < 15]

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
6	Male	Loyal Customer	10	Personal Travel	Eco	1812	0	0.000	1	0	...	2	3.000	3	4	5	4	2	4.000	2.000	2.000
13	Female	Loyal Customer	13	Personal Travel	Eco	3693	5	0.000	1	0	...	4	4.000	4	1	3	1	4	2.000	4.000	1.000
16	Female	Loyal Customer	9	Personal Travel	Eco	3305	0	0.000	1	0	...	3	1.000	1	1	3	3	3	2.000	4.000	1.000
17	Female	Loyal Customer	10	Personal Travel	Eco	2090	0	0.000	1	0	...	1	3.000	5	1	4	2	1	2.000	1.000	1.000
24	Male	Loyal Customer	9	Personal Travel	Eco	972	0	0.000	1	0	...	4	4.000	3	3	1	3	4	2.000	4.000	2.000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
90881	Female	disloyal Customer	12	Personal Travel	Eco	2429	0	3.000	0	3	...	1	2.000	4	1	3	4	1	2.000	1.000	3.000
90893	Female	disloyal Customer	7	Personal Travel	Eco	1616	0	15.000	0	3	...	3	NaN	1	4	3	1	3	3.000	3.000	3.000
90900	Female	disloyal Customer	14	Personal Travel	Business	1966	0	0.000	0	3	...	1	5.000	5	4	4	4	1	4.000	1.000	3.000
90901	Female	disloyal Customer	14	Personal Travel	Eco	1972	0	0.000	0	3	...	5	3.000	4	5	4	4	5	4.000	5.000	4.000
90912	Female	disloyal Customer	11	Personal Travel	Eco	2752	5	0.000	1	5	...	2	3.000	5	3	5	4	2	3.000	2.000	4.000

4488 rows × 26 columns

• There's 4488 customers under 15 years old that filled out the survey with wrong info so I will put them as a missing value and impute after.

df.loc[(df.Age < 15), "Age"] = np.nan

# check if missing values for Age are in place
df.info()

<class 'pandas.core.frame.DataFrame'>
Int64Index: 90917 entries, 0 to 90916
Data columns (total 26 columns):
 #   Column                             Non-Null Count  Dtype   
---  ------                             --------------  -----   
 0   Gender                             90917 non-null  category
 1   CustomerType                       81818 non-null  category
 2   Age                                86429 non-null  float64 
 3   TypeTravel                         81829 non-null  category
 4   Class                              90917 non-null  category
 5   Flight_Distance                    90917 non-null  int64   
 6   DepartureDelayin_Mins              90917 non-null  int64   
 7   ArrivalDelayin_Mins                90633 non-null  float64 
 8   Satisfaction                       90917 non-null  int64   
 9   Seat_comfort                       90917 non-null  int64   
 10  Departure.Arrival.time_convenient  82673 non-null  float64 
 11  Food_drink                         82736 non-null  float64 
 12  Gate_location                      90917 non-null  int64   
 13  Inflightwifi_service               90917 non-null  int64   
 14  Inflight_entertainment             90917 non-null  int64   
 15  Online_support                     90917 non-null  int64   
 16  Ease_of_Onlinebooking              90917 non-null  int64   
 17  Onboard_service                    83738 non-null  float64 
 18  Leg_room_service                   90917 non-null  int64   
 19  Baggage_handling                   90917 non-null  int64   
 20  Checkin_service                    90917 non-null  int64   
 21  Cleanliness                        90917 non-null  int64   
 22  Online_boarding                    90917 non-null  int64   
 23  Facilities_avg                     90917 non-null  float64 
 24  Online_service_avg                 90917 non-null  float64 
 25  InFlight_avg                       90917 non-null  float64 
dtypes: category(4), float64(8), int64(14)
memory usage: 16.3 MB

• Checked!

Observations from Flight Distance

histogram_boxplot(df, "Flight_Distance")

• Plot suggest a normal distribution, however, there is a peak between 200-350.
• 75% of the travelers use this line to traveldistance less than 2542
• Boxplot is showing outliers in the upper whisker in ~4400

df[df["Flight_Distance"] > 4400]

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
105	Female	Loyal Customer	21.000	Personal Travel	Eco	4804	20	NaN	1	0	...	5	NaN	1	4	2	3	5	3.000	5.000	2.000
400	Female	Loyal Customer	63.000	Personal Travel	Business	6591	3	0.000	1	0	...	3	2.000	4	3	1	5	1	2.000	2.000	3.000
801	Male	Loyal Customer	35.000	Personal Travel	Eco	6470	0	32.000	0	1	...	1	2.000	2	2	3	2	3	2.000	2.000	2.000
802	Male	Loyal Customer	23.000	Personal Travel	Eco	4650	0	NaN	0	1	...	2	1.000	3	4	4	4	2	3.000	2.000	2.000
1156	Male	Loyal Customer	21.000	Personal Travel	Eco	5127	2	2.000	0	1	...	3	5.000	1	2	2	2	2	3.000	2.000	1.000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
90432	Female	NaN	31.000	Business travel	Business	4889	19	14.000	1	1	...	5	5.000	3	5	4	4	5	4.000	5.000	3.000
90548	Male	Loyal Customer	34.000	Business travel	Eco	5832	4	1.000	1	5	...	4	5.000	2	5	5	5	5	5.000	5.000	4.000
90672	Female	Loyal Customer	22.000	NaN	Business	4652	0	0.000	1	3	...	5	4.000	3	4	3	5	5	4.000	5.000	4.000
90710	Female	NaN	42.000	Business travel	Business	5403	189	153.000	1	1	...	5	5.000	5	4	5	4	5	4.000	5.000	3.000
90843	Female	disloyal Customer	15.000	Personal Travel	Eco	4522	0	0.000	0	1	...	1	5.000	3	5	5	5	1	5.000	1.000	2.000

1623 rows × 26 columns

• We can see that there are 1623 observations which can be considered as outliers.
• We will not treat these outliers as they represent data

Observations from Departure Delay in minutes

histogram_boxplot(df, "DepartureDelayin_Mins")

• Most of the travelers had zero mins of delays, skewing the distibution toward zero (right).
• Boxplot is showing outliers in the upper whisker after ~35

df[df["DepartureDelayin_Mins"] > 35]

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
8	Female	Loyal Customer	58.000	Personal Travel	Eco	104	47	48.000	1	0	...	3	3.000	0	1	2	3	5	2.000	4.000	1.000
11	Female	Loyal Customer	47.000	Personal Travel	Eco	84	40	48.000	1	0	...	5	5.000	0	5	2	5	2	3.000	3.000	2.000
36	Female	Loyal Customer	17.000	Personal Travel	Eco	2748	427	440.000	1	0	...	3	1.000	3	4	4	1	4	3.000	4.000	1.000
110	Female	Loyal Customer	55.000	Personal Travel	Eco	994	68	80.000	1	0	...	3	3.000	0	3	1	3	3	3.000	3.000	2.000
128	Male	Loyal Customer	24.000	Personal Travel	Eco	3219	93	86.000	1	0	...	3	4.000	2	5	5	5	3	4.000	3.000	2.000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
90892	Female	disloyal Customer	54.000	Personal Travel	Eco	2928	41	41.000	0	3	...	1	4.000	2	1	2	1	1	2.000	1.000	3.000
90910	Female	NaN	70.000	Personal Travel	Eco	1674	54	46.000	1	5	...	5	3.000	2	4	5	4	5	3.000	5.000	4.000
90914	Male	disloyal Customer	69.000	Personal Travel	Eco	2320	155	163.000	0	3	...	4	4.000	3	4	2	3	2	3.000	3.000	3.000
90915	Male	disloyal Customer	66.000	Personal Travel	Eco	2450	193	205.000	0	3	...	3	3.000	2	3	2	1	2	2.000	2.000	2.000
90916	Female	disloyal Customer	38.000	Personal Travel	Eco	4307	185	186.000	0	3	...	4	5.000	5	5	3	3	3	4.000	3.000	3.000

11053 rows × 26 columns

• We can see that there are 11053 observations which can be considered as outliers.
• We will not treat these outliers as they represent data

Observations from Arrival Delay in minutes

histogram_boxplot(df, "ArrivalDelayin_Mins")

• Most of the travelers had zero mins of delays, skewing the distibution toward zero (right).
• Boxplot is showing outliers in the upper whisker after ~35

df[df["ArrivalDelayin_Mins"] > 35]

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
8	Female	Loyal Customer	58.000	Personal Travel	Eco	104	47	48.000	1	0	...	3	3.000	0	1	2	3	5	2.000	4.000	1.000
11	Female	Loyal Customer	47.000	Personal Travel	Eco	84	40	48.000	1	0	...	5	5.000	0	5	2	5	2	3.000	3.000	2.000
36	Female	Loyal Customer	17.000	Personal Travel	Eco	2748	427	440.000	1	0	...	3	1.000	3	4	4	1	4	3.000	4.000	1.000
61	Female	NaN	67.000	Personal Travel	Eco	465	35	69.000	1	0	...	4	4.000	0	2	1	4	3	3.000	4.000	1.000
110	Female	Loyal Customer	55.000	Personal Travel	Eco	994	68	80.000	1	0	...	3	3.000	0	3	1	3	3	3.000	3.000	2.000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
90894	Female	disloyal Customer	20.000	Personal Travel	Business	2946	8	37.000	0	3	...	3	3.000	3	5	3	5	3	3.000	3.000	4.000
90910	Female	NaN	70.000	Personal Travel	Eco	1674	54	46.000	1	5	...	5	3.000	2	4	5	4	5	3.000	5.000	4.000
90914	Male	disloyal Customer	69.000	Personal Travel	Eco	2320	155	163.000	0	3	...	4	4.000	3	4	2	3	2	3.000	3.000	3.000
90915	Male	disloyal Customer	66.000	Personal Travel	Eco	2450	193	205.000	0	3	...	3	3.000	2	3	2	1	2	2.000	2.000	2.000
90916	Female	disloyal Customer	38.000	Personal Travel	Eco	4307	185	186.000	0	3	...	4	5.000	5	5	3	3	3	4.000	3.000	3.000

11241 rows × 26 columns

• We can see that there are 11241 observations which can be considered as outliers.
• We will not treat these outliers as they represent data

EDA for Categorical Variables

# function to create labeled barplots


def labeled_barplot(data, feature, perc=False, n=None):
    """
    Barplot with percentage at the top

    data: dataframe
    feature: dataframe column
    perc: whether to display percentages instead of count (default is False)
    n: displays the top n category levels (default is None, i.e., display all levels)
    """

    total = len(data[feature])  # length of the column
    count = data[feature].nunique()
    if n is None:
        plt.figure(figsize=(count + 1, 5))
    else:
        plt.figure(figsize=(n + 1, 5))

    plt.xticks(rotation=90, fontsize=15)
    ax = sns.countplot(
        data=data,
        x=feature,
        palette="Paired",
        order=data[feature].value_counts().index[:n].sort_values(),
    )

    for p in ax.patches:
        if perc == True:
            label = "{:.1f}%".format(
                100 * p.get_height() / total
            )  # percentage of each class of the category
        else:
            label = p.get_height()  # count of each level of the category

        x = p.get_x() + p.get_width() / 2  # width of the plot
        y = p.get_height()  # height of the plot

        ax.annotate(
            label,
            (x, y),
            ha="center",
            va="center",
            size=12,
            xytext=(0, 5),
            textcoords="offset points",
        )  # annotate the percentage

    plt.show()  # show the plot

Observation from Gender

labeled_barplot(df, "Gender")

• Female and male are practically distributed even for this dataset

Observation from Customer Type

labeled_barplot(df, "CustomerType")

• Most of the customer in this dataset are loyal customers in a 73.58%

Observation from Type Travel

labeled_barplot(df, "TypeTravel")

• Most of customers from this dataset have a business flight as a purpose in 62.12%

Observations from Class

labeled_barplot(df, "Class")

• 47.88% of the travelers where in business class, follow by economical class in 44.82%
• Let's check attributes for each class

# Grouping customers for class
df_eco = df[df["Class"] == "Eco"]
df_ecoplus = df[df["Class"] == "Eco Plus"]
df_bus = df[df["Class"] == "Business"]

df_eco.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	40758	2	Female	20690	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	36590	2	Loyal Customer	28062	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	37415.000	NaN	NaN	NaN	39.667	15.401	15.000	26.000	38.000	52.000	85.000
TypeTravel	36674	2	Personal Travel	20738	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	40758	1	Eco	40758	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	40758.000	NaN	NaN	NaN	1824.298	794.114	50.000	1379.000	1841.000	2277.000	6924.000
DepartureDelayin_Mins	40758.000	NaN	NaN	NaN	15.023	39.411	0.000	0.000	0.000	12.000	1592.000
ArrivalDelayin_Mins	40619.000	NaN	NaN	NaN	15.554	39.849	0.000	0.000	0.000	14.000	1584.000
Satisfaction	40758.000	NaN	NaN	NaN	0.393	0.488	0.000	0.000	0.000	1.000	1.000
Seat_comfort	40758.000	NaN	NaN	NaN	2.875	1.347	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	37121.000	NaN	NaN	NaN	3.075	1.547	0.000	2.000	3.000	4.000	5.000
Food_drink	37140.000	NaN	NaN	NaN	2.776	1.399	0.000	2.000	3.000	4.000	5.000
Gate_location	40758.000	NaN	NaN	NaN	3.000	1.239	0.000	2.000	3.000	4.000	5.000
Inflightwifi_service	40758.000	NaN	NaN	NaN	3.166	1.353	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	40758.000	NaN	NaN	NaN	3.057	1.378	0.000	2.000	3.000	4.000	5.000
Online_support	40758.000	NaN	NaN	NaN	3.287	1.371	1.000	2.000	4.000	4.000	5.000
Ease_of_Onlinebooking	40758.000	NaN	NaN	NaN	3.306	1.352	0.000	2.000	4.000	4.000	5.000
Onboard_service	37578.000	NaN	NaN	NaN	3.273	1.308	1.000	2.000	3.000	4.000	5.000
Leg_room_service	40758.000	NaN	NaN	NaN	3.328	1.335	0.000	2.000	4.000	4.000	5.000
Baggage_handling	40758.000	NaN	NaN	NaN	3.569	1.171	1.000	3.000	4.000	4.000	5.000
Checkin_service	40758.000	NaN	NaN	NaN	3.185	1.299	1.000	2.000	3.000	4.000	5.000
Cleanliness	40758.000	NaN	NaN	NaN	3.583	1.165	1.000	3.000	4.000	4.000	5.000
Online_boarding	40758.000	NaN	NaN	NaN	3.228	1.347	0.000	2.000	3.000	4.000	5.000
Facilities_avg	40758.000	NaN	NaN	NaN	3.262	0.814	1.000	3.000	3.000	4.000	5.000
Online_service_avg	40758.000	NaN	NaN	NaN	3.275	1.243	1.000	2.000	3.000	4.000	5.000
InFlight_avg	40758.000	NaN	NaN	NaN	3.129	0.804	0.000	3.000	3.000	4.000	5.000

df_ecoplus.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	6624	2	Female	3509	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	5941	2	Loyal Customer	5363	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	6130.000	NaN	NaN	NaN	41.027	15.068	15.000	29.000	39.000	53.000	85.000
TypeTravel	5982	2	Business travel	3104	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	6624	1	Eco Plus	6624	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	6624.000	NaN	NaN	NaN	1790.261	832.067	50.000	1326.000	1804.000	2261.000	6889.000
DepartureDelayin_Mins	6624.000	NaN	NaN	NaN	15.174	36.029	0.000	0.000	0.000	13.000	469.000
ArrivalDelayin_Mins	6601.000	NaN	NaN	NaN	15.613	36.138	0.000	0.000	0.000	15.000	518.000
Satisfaction	6624.000	NaN	NaN	NaN	0.426	0.495	0.000	0.000	0.000	1.000	1.000
Seat_comfort	6624.000	NaN	NaN	NaN	2.931	1.384	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	6039.000	NaN	NaN	NaN	3.106	1.487	0.000	2.000	3.000	4.000	5.000
Food_drink	6038.000	NaN	NaN	NaN	2.810	1.441	0.000	2.000	3.000	4.000	5.000
Gate_location	6624.000	NaN	NaN	NaN	2.982	1.296	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	6624.000	NaN	NaN	NaN	3.196	1.351	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	6624.000	NaN	NaN	NaN	3.097	1.382	0.000	2.000	3.000	4.000	5.000
Online_support	6624.000	NaN	NaN	NaN	3.298	1.349	1.000	2.000	4.000	4.000	5.000
Ease_of_Onlinebooking	6624.000	NaN	NaN	NaN	3.323	1.342	0.000	2.000	4.000	4.000	5.000
Onboard_service	6091.000	NaN	NaN	NaN	3.168	1.326	1.000	2.000	3.000	4.000	5.000
Leg_room_service	6624.000	NaN	NaN	NaN	3.284	1.355	0.000	2.000	3.000	4.000	5.000
Baggage_handling	6624.000	NaN	NaN	NaN	3.462	1.182	1.000	3.000	4.000	4.000	5.000
Checkin_service	6624.000	NaN	NaN	NaN	3.077	1.323	1.000	2.000	3.000	4.000	5.000
Cleanliness	6624.000	NaN	NaN	NaN	3.492	1.160	1.000	3.000	4.000	4.000	5.000
Online_boarding	6624.000	NaN	NaN	NaN	3.220	1.337	0.000	2.000	3.000	4.000	5.000
Facilities_avg	6624.000	NaN	NaN	NaN	3.175	0.820	1.000	3.000	3.000	4.000	5.000
Online_service_avg	6624.000	NaN	NaN	NaN	3.283	1.222	1.000	2.000	3.000	4.000	5.000
InFlight_avg	6624.000	NaN	NaN	NaN	3.137	0.810	0.000	3.000	3.000	4.000	5.000

df_bus.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	43535	2	Female	21987	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	39287	2	Loyal Customer	33472	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	42884.000	NaN	NaN	NaN	42.049	12.294	15.000	33.000	42.000	51.000	85.000
TypeTravel	39173	2	Business travel	37441	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	43535	1	Business	43535	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	43535.000	NaN	NaN	NaN	2158.043	1202.562	50.000	1326.000	2101.000	3001.000	6950.000
DepartureDelayin_Mins	43535.000	NaN	NaN	NaN	14.298	38.352	0.000	0.000	0.000	12.000	1305.000
ArrivalDelayin_Mins	43413.000	NaN	NaN	NaN	14.511	38.689	0.000	0.000	0.000	12.000	1280.000
Satisfaction	43535.000	NaN	NaN	NaN	0.711	0.453	0.000	0.000	1.000	1.000	1.000
Seat_comfort	43535.000	NaN	NaN	NaN	2.791	1.435	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	39513.000	NaN	NaN	NaN	2.899	1.504	0.000	2.000	3.000	4.000	5.000
Food_drink	39558.000	NaN	NaN	NaN	2.925	1.479	0.000	2.000	3.000	4.000	5.000
Gate_location	43535.000	NaN	NaN	NaN	2.983	1.371	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	43535.000	NaN	NaN	NaN	3.340	1.278	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	43535.000	NaN	NaN	NaN	3.734	1.207	0.000	3.000	4.000	5.000	5.000
Online_support	43535.000	NaN	NaN	NaN	3.770	1.190	0.000	3.000	4.000	5.000	5.000
Ease_of_Onlinebooking	43535.000	NaN	NaN	NaN	3.657	1.227	0.000	3.000	4.000	5.000	5.000
Onboard_service	40069.000	NaN	NaN	NaN	3.693	1.181	0.000	3.000	4.000	5.000	5.000
Leg_room_service	43535.000	NaN	NaN	NaN	3.667	1.215	0.000	3.000	4.000	5.000	5.000
Baggage_handling	43535.000	NaN	NaN	NaN	3.853	1.113	1.000	3.000	4.000	5.000	5.000
Checkin_service	43535.000	NaN	NaN	NaN	3.526	1.185	0.000	3.000	4.000	4.000	5.000
Cleanliness	43535.000	NaN	NaN	NaN	3.857	1.111	0.000	3.000	4.000	5.000	5.000
Online_boarding	43535.000	NaN	NaN	NaN	3.489	1.233	0.000	3.000	4.000	4.000	5.000
Facilities_avg	43535.000	NaN	NaN	NaN	3.514	0.793	1.000	3.000	4.000	4.000	5.000
Online_service_avg	43535.000	NaN	NaN	NaN	3.638	1.046	1.000	3.000	4.000	4.000	5.000
InFlight_avg	43535.000	NaN	NaN	NaN	3.332	0.814	0.000	3.000	3.000	4.000	5.000

• Apperantly Age, Gender. Customer Type, Flight distance and delays are fairly distributed in the same way for the 3 three classes.
• Most of the customers that travel in economic class has a personal purpose, while Eco Plus and Business class the main purpose is business.
• Most of the Business class travelers are satisfied with the airline service, while Eco and EcoPlus showed to be neutral or dissatisfied.
• Apperantly, The check in service, online boarding and onboard service weight and make the difference when the traveler is satisfied or not with the airline service.

Observations from Satisfaction

labeled_barplot(df, "Satisfaction")

• 54.73% of the travelers are satisfied with the airline services overall.

Observations from Seat Comfort

labeled_barplot(df, "Seat_comfort")

• Most of the data is split in needs improvement 22% (2), acceptable 22.6% (3) and good 21.76% (4).

Observations from Departure Arrival Time Convinient

labeled_barplot(df, "Departure.Arrival.time_convenient")

• 20.72% of the travelers consider that the Departure and arrival time are convinient.
• Probably the 19.14% that consider the Departure/arrival time convinient as extremely poor or poor belong to the arrival/departure outliers area.

Observations from Food drink

labeled_barplot(df, "Food_drink")

• Most of the data is split in needs improvement 19.09% (2), acceptable 19.77% (3) and good 18.95% (4)

Observations from Gate Location

labeled_barplot(df, "Gate_location")

• Most of the data is split in needs improvement 18.81% (2), manageable 25.70% (3) and convinient 23.18% (4)

Observations from In Flight wifi service

labeled_barplot(df, "Inflightwifi_service")

• Most of the data is split in needs improvement 20.77% (2), acceptable 21.11% (3), good 24.36% (4) and excellent 22.26% (5)

Observations from Inflight entertaiment

labeled_barplot(df, "Inflight_entertainment")

• 32.29% of travelers says that the in flight enterteiment was good.

Observations from Online Support

labeled_barplot(df, "Online_support")

• 31.92% of travelers think that the online support is good, followed by 27.39% that thinks that it is excellent.

Observations from Ease Online booking

labeled_barplot(df, "Ease_of_Onlinebooking")

• Same tendency as online support 30.77% of the travelers think is good followed by 26.33% that thinks is excellent.

Observation from Onboard service

labeled_barplot(df, "Onboard_service")

• 29.99% of the travellers think that the onboard service is good.

Observations from Leg room services

labeled_barplot(df, "Leg_room_service")

• 30.57% of the travellers think that the leg room service is good, followed by 26.46% that thinks is excellent.

Observations from baggage handling

labeled_barplot(df, "Baggage_handling")

• 37.18% of the travellers think that the baggage handling is good followed by 27.48% that thinks is excellent.

Observations from Check in Service

labeled_barplot(df, "Checkin_service")

• 28.01% of the travelers think that the check in service is good, followed by 27.42% that think is acceptable.

Observations from Cleanliness

labeled_barplot(df, "Cleanliness")

• 37.64% of the travelers think that the cleanliness is good.

Observations from Online Boarding

labeled_barplot(df, "Online_boarding")

• 27.12% of the travelers think that it is good, followed by 23.55% that say is acceptable and 23.08% that say is excellent

Observations from Facilities Average

labeled_barplot(df, "Facilities_avg")

• 44089 Think that the facilities on average are good, however a good enough proportion of the passangers think the service is acceptable and needs improvement

Observations from Online Service

labeled_barplot(df, "Online_service_avg")

• 33877 passangers think that the online services are good, however theres a good proportion of the passangersthink that is acceptable and needs improvement.

Observations from Inflight Services

labeled_barplot(df, "InFlight_avg")

• 39594 passangers think that the inflight services is acceptable, however theres a good proportion that think it is good and others that needs improvement.

Bivariate Analysis

Corralation Matrix

corr_matrix = df.corr()
corr_matrix

	Age	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	Departure.Arrival.time_convenient	Food_drink	Gate_location	Inflightwifi_service	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
Age	1.000	-0.256	-0.007	-0.010	0.105	0.009	0.065	0.015	-0.002	0.007	...	0.070	0.070	0.092	-0.009	0.032	-0.011	0.031	0.030	0.080	0.073
Flight_Distance	-0.256	1.000	0.111	0.108	-0.040	-0.046	-0.001	-0.009	-0.003	0.012	...	-0.024	-0.034	-0.033	0.017	0.004	0.007	0.010	-0.004	-0.017	-0.022
DepartureDelayin_Mins	-0.007	0.111	1.000	0.966	-0.073	-0.022	0.005	-0.013	0.006	-0.033	...	-0.036	-0.039	0.002	-0.011	-0.021	-0.062	-0.019	-0.023	-0.032	-0.038
ArrivalDelayin_Mins	-0.010	0.108	0.966	1.000	-0.079	-0.024	0.003	-0.015	0.006	-0.035	...	-0.039	-0.041	-0.000	-0.015	-0.023	-0.067	-0.021	-0.027	-0.035	-0.042
Satisfaction	0.105	-0.040	-0.073	-0.079	1.000	0.241	-0.015	0.118	-0.012	0.226	...	0.430	0.351	0.305	0.259	0.267	0.259	0.335	0.327	0.430	0.392
Seat_comfort	0.009	-0.046	-0.022	-0.024	0.241	1.000	0.435	0.719	0.408	0.130	...	0.210	0.120	0.137	0.118	0.045	0.109	0.131	0.245	0.173	0.712
Departure.Arrival.time_convenient	0.065	-0.001	0.005	0.003	-0.015	0.435	1.000	0.527	0.546	-0.005	...	0.000	0.060	0.027	0.068	0.063	0.064	0.000	0.277	0.001	0.535
Food_drink	0.015	-0.009	-0.013	-0.015	0.118	0.719	0.527	1.000	0.526	0.027	...	0.041	0.039	0.075	0.037	0.015	0.033	0.015	0.229	0.033	0.676
Gate_location	-0.002	-0.003	0.006	0.006	-0.012	0.408	0.546	0.526	1.000	-0.004	...	0.001	-0.024	-0.007	-0.001	-0.030	-0.002	-0.002	0.350	-0.000	0.351
Inflightwifi_service	0.007	0.012	-0.033	-0.035	0.226	0.130	-0.005	0.027	-0.004	1.000	...	0.603	0.058	0.035	0.041	0.091	0.039	0.631	0.071	0.668	0.356
Inflight_entertainment	0.126	-0.029	-0.029	-0.032	0.522	0.423	0.077	0.365	0.000	0.251	...	0.319	0.181	0.160	0.115	0.226	0.109	0.354	0.197	0.416	0.576
Online_support	0.116	-0.031	-0.033	-0.035	0.387	0.122	0.002	0.032	0.004	0.557	...	0.617	0.156	0.141	0.103	0.204	0.096	0.671	0.177	0.852	0.333
Ease_of_Onlinebooking	0.070	-0.024	-0.036	-0.039	0.430	0.210	0.000	0.041	0.001	0.603	...	1.000	0.432	0.355	0.399	0.137	0.418	0.684	0.356	0.859	0.455
Onboard_service	0.070	-0.034	-0.039	-0.041	0.351	0.120	0.060	0.039	-0.024	0.058	...	0.432	1.000	0.409	0.527	0.249	0.549	0.135	0.685	0.269	0.326
Leg_room_service	0.092	-0.033	0.002	-0.000	0.305	0.137	0.027	0.075	-0.007	0.035	...	0.355	0.409	1.000	0.408	0.169	0.411	0.114	0.364	0.227	0.427
Baggage_handling	-0.009	0.017	-0.011	-0.015	0.259	0.118	0.068	0.037	-0.001	0.041	...	0.399	0.527	0.408	1.000	0.241	0.632	0.114	0.665	0.230	0.322
Checkin_service	0.032	0.004	-0.021	-0.023	0.267	0.045	0.063	0.015	-0.030	0.091	...	0.137	0.249	0.169	0.241	1.000	0.242	0.183	0.571	0.196	0.198
Cleanliness	-0.011	0.007	-0.062	-0.067	0.259	0.109	0.064	0.033	-0.002	0.039	...	0.418	0.549	0.411	0.632	0.242	1.000	0.106	0.535	0.231	0.393
Online_boarding	0.031	0.010	-0.019	-0.021	0.335	0.131	0.000	0.015	-0.002	0.631	...	0.684	0.135	0.114	0.114	0.183	0.106	1.000	0.162	0.883	0.323
Facilities_avg	0.030	-0.004	-0.023	-0.027	0.327	0.245	0.277	0.229	0.350	0.071	...	0.356	0.685	0.364	0.665	0.571	0.535	0.162	1.000	0.259	0.446
Online_service_avg	0.080	-0.017	-0.032	-0.035	0.430	0.173	0.001	0.033	-0.000	0.668	...	0.859	0.269	0.227	0.230	0.196	0.231	0.883	0.259	1.000	0.414
InFlight_avg	0.073	-0.022	-0.038	-0.042	0.392	0.712	0.535	0.676	0.351	0.356	...	0.455	0.326	0.427	0.322	0.198	0.393	0.323	0.446	0.414	1.000

22 rows × 22 columns

sns.set(style="white", font_scale=2.2)
fig = plt.figure(figsize=[35, 30])
mask = np.triu(np.ones_like(corr_matrix, dtype=bool))
cmap = sns.diverging_palette(150, 0, as_cmap=True)
sns.heatmap(
    corr_matrix,
    cmap="seismic",
    linewidth=3,
    linecolor="white",
    vmax=1,
    vmin=-1,
    mask=mask,
    annot=True,
    fmt="0.2f",
)
plt.title("Correlation Heatmap", weight="bold", fontsize=50)
plt.savefig("heatmap.png", transparent=True, bbox_inches="tight")

Observations

• The target variable (Satisfaction)has a very low correlation (less than 0.11) with features like Age, Flight Distance, Departure Delay in minutes, Arrival Delay in Minutes, Arrival time conveinient and gate Location
• There is a strong linear positive correlation between Departure Delay in Mins and Arrival Delay in Mins, indicating that both variables are proportional to each other.
• There is a positive correlation between Seat Comfort and Food Drink.
• There is a positive correlation between Online Boarding and Online Support
• There is a positive correlation between Online Boarding and Ease of Online Booking.
• There's a high relationship between Facilities Average with onboard services, probably the most important that belongs under this parameter.
• Online Service Average, have a highly positive relationship with its three components, Online Support, Ease online booking and Online Onboarding.
• Inflight Service Average have a high correlation with seat confort and food/drink escentially.

sns.pairplot(data=df)

<seaborn.axisgrid.PairGrid at 0x28fc7362400>

Observations

• There is a perfectly linear relationship between Departure Delay in Mins and Arrival Delay in Mins

# function to plot stacked bar chart


def stacked_barplot(df, predictor, target):
    """
    Print the category counts and plot a stacked bar chart

    data: dataframe
    predictor: independent variable
    target: target variable
    """
    count = df[predictor].nunique()
    sorter = df[target].value_counts().index[-1]
    tab1 = pd.crosstab(df[predictor], df[target], margins=True).sort_values(
        by=sorter, ascending=False
    )
    print(tab1)
    print("-" * 120)
    tab = pd.crosstab(df[predictor], df[target], normalize="index").sort_values(
        by=sorter, ascending=False
    )
    tab.plot(kind="bar", stacked=True, figsize=(count + 1, 5))
    plt.legend(
        loc="lower left",
        frameon=False,
    )
    plt.legend(loc="upper left", bbox_to_anchor=(1, 1))
    plt.show()

Bivariate Analysis for categorical variables

Satisfaction vs Gender

stacked_barplot(df, "Gender", "Satisfaction")

Satisfaction      0      1    All
Gender                           
All           41156  49761  90917
Male          25025  19706  44731
Female        16131  30055  46186
------------------------------------------------------------------------------------------------------------------------

• Females tend to have more satisfaction with the airline service with a 65% while males are 44.05%

Satisfaction vs Customer Type

stacked_barplot(df, "CustomerType", "Satisfaction")

Satisfaction           0      1    All
CustomerType                          
All                36931  44887  81818
Loyal Customer     25580  41317  66897
disloyal Customer  11351   3570  14921
------------------------------------------------------------------------------------------------------------------------

• Loyal Customers tend to be more satisfied in 61.76% than disloyal customers in 23.92%

Let's check why Loyals and disloyal customers are not satisfied

df_loyal = df.query('CustomerType == "Loyal Customer" & Satisfaction==0')

df_loyal.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	25580	2	Male	17352	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	25580	1	Loyal Customer	25580	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	23601.000	NaN	NaN	NaN	42.773	15.089	15.000	31.000	43.000	55.000	85.000
TypeTravel	22772	2	Personal Travel	11978	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	25580	3	Eco	14867	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	25580.000	NaN	NaN	NaN	2032.279	986.980	50.000	1463.750	1964.000	2539.000	6924.000
DepartureDelayin_Mins	25580.000	NaN	NaN	NaN	18.804	47.194	0.000	0.000	0.000	17.000	1592.000
ArrivalDelayin_Mins	25495.000	NaN	NaN	NaN	19.470	47.407	0.000	0.000	0.000	18.000	1584.000
Satisfaction	25580.000	NaN	NaN	NaN	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Seat_comfort	25580.000	NaN	NaN	NaN	2.498	1.009	0.000	2.000	3.000	3.000	5.000
Departure.Arrival.time_convenient	23311.000	NaN	NaN	NaN	3.338	1.450	0.000	2.000	4.000	5.000	5.000
Food_drink	23260.000	NaN	NaN	NaN	2.777	1.313	0.000	2.000	3.000	4.000	5.000
Gate_location	25580.000	NaN	NaN	NaN	2.993	1.274	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	25580.000	NaN	NaN	NaN	2.878	1.328	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	25580.000	NaN	NaN	NaN	2.688	1.108	0.000	2.000	3.000	4.000	5.000
Online_support	25580.000	NaN	NaN	NaN	2.942	1.224	0.000	2.000	3.000	4.000	5.000
Ease_of_Onlinebooking	25580.000	NaN	NaN	NaN	2.774	1.257	0.000	2.000	3.000	4.000	5.000
Onboard_service	23584.000	NaN	NaN	NaN	2.931	1.255	0.000	2.000	3.000	4.000	5.000
Leg_room_service	25580.000	NaN	NaN	NaN	2.997	1.280	0.000	2.000	3.000	4.000	5.000
Baggage_handling	25580.000	NaN	NaN	NaN	3.283	1.153	1.000	3.000	3.000	4.000	5.000
Checkin_service	25580.000	NaN	NaN	NaN	2.931	1.270	0.000	2.000	3.000	4.000	5.000
Cleanliness	25580.000	NaN	NaN	NaN	3.301	1.152	0.000	3.000	3.000	4.000	5.000
Online_boarding	25580.000	NaN	NaN	NaN	2.790	1.280	0.000	2.000	3.000	4.000	5.000
Facilities_avg	25580.000	NaN	NaN	NaN	3.035	0.830	1.000	2.000	3.000	4.000	5.000
Online_service_avg	25580.000	NaN	NaN	NaN	2.836	1.123	1.000	2.000	3.000	4.000	5.000
InFlight_avg	25580.000	NaN	NaN	NaN	2.924	0.703	1.000	2.000	3.000	3.000	5.000

df_disloyal = df.query('CustomerType == "disloyal Customer" & Satisfaction==0')

df_disloyal.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	11351	2	Female	6248	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	11351	1	disloyal Customer	11351	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	10969.000	NaN	NaN	NaN	31.788	10.946	15.000	24.000	28.000	38.000	85.000
TypeTravel	10065	2	Business travel	9972	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	11351	3	Eco	7314	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	11351.000	NaN	NaN	NaN	2021.777	601.981	220.000	1613.500	1952.000	2330.000	6837.000
DepartureDelayin_Mins	11351.000	NaN	NaN	NaN	16.008	38.108	0.000	0.000	0.000	15.000	569.000
ArrivalDelayin_Mins	11316.000	NaN	NaN	NaN	16.639	38.688	0.000	0.000	0.000	16.000	600.000
Satisfaction	11351.000	NaN	NaN	NaN	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Seat_comfort	11351.000	NaN	NaN	NaN	2.403	0.950	1.000	2.000	2.000	3.000	4.000
Departure.Arrival.time_convenient	10330.000	NaN	NaN	NaN	2.297	1.371	0.000	1.000	2.000	3.000	5.000
Food_drink	10333.000	NaN	NaN	NaN	2.409	1.015	0.000	2.000	2.000	3.000	5.000
Gate_location	11351.000	NaN	NaN	NaN	3.043	1.069	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	11351.000	NaN	NaN	NaN	3.029	1.389	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	11351.000	NaN	NaN	NaN	2.443	1.040	0.000	2.000	2.000	3.000	5.000
Online_support	11351.000	NaN	NaN	NaN	3.011	1.413	1.000	2.000	3.000	4.000	5.000
Ease_of_Onlinebooking	11351.000	NaN	NaN	NaN	3.052	1.391	1.000	2.000	3.000	4.000	5.000
Onboard_service	10436.000	NaN	NaN	NaN	3.089	1.284	1.000	2.000	3.000	4.000	5.000
Leg_room_service	11351.000	NaN	NaN	NaN	3.179	1.338	1.000	2.000	3.000	4.000	5.000
Baggage_handling	11351.000	NaN	NaN	NaN	3.565	1.073	1.000	3.000	4.000	4.000	5.000
Checkin_service	11351.000	NaN	NaN	NaN	3.065	1.296	1.000	2.000	3.000	4.000	5.000
Cleanliness	11351.000	NaN	NaN	NaN	3.572	1.088	1.000	3.000	4.000	4.000	5.000
Online_boarding	11351.000	NaN	NaN	NaN	3.064	1.394	1.000	2.000	3.000	4.000	5.000
Facilities_avg	11351.000	NaN	NaN	NaN	3.199	0.779	1.000	3.000	3.000	4.000	5.000
Online_service_avg	11351.000	NaN	NaN	NaN	3.043	1.334	1.000	2.000	3.000	4.000	5.000
InFlight_avg	11351.000	NaN	NaN	NaN	2.768	0.697	1.000	2.000	3.000	3.000	5.000

• Loyal customers that are neutral or not satisfied with the airline services, most of them are males that travel for personal reasons in Eco class tend to think on average that except for Departure arrival time convinient is good, the other parameters are acceptable.
• disloyal customers that are neutral or not satisfied with the airline services, most of them are females that travel for business purposed in Eco class that the airline needs improvement on average for seat confort, departure arrival time convenient, food and drink and In flight enterteiment.

Satisfaction vs Type Travel

stacked_barplot(df, "TypeTravel", "Satisfaction")

Satisfaction         0      1    All
TypeTravel                          
All              37062  44767  81829
Business travel  23490  32991  56481
Personal Travel  13572  11776  25348
------------------------------------------------------------------------------------------------------------------------

• Business travellers are more satisfied in a 58.41% with the airline services than the ones who travel for personal reasons 46.45%.

Let's check why customers are neutral or insatisfied with the airline service by Type Travel.

df_PT = df.query('TypeTravel == "Personal Travel" & Satisfaction==0')
df_PT.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	13572	2	Male	11441	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	12071	2	Loyal Customer	11978	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	11909.000	NaN	NaN	NaN	42.591	16.192	15.000	29.000	43.000	57.000	70.000
TypeTravel	13572	1	Personal Travel	13572	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	13572	3	Eco	11112	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	13572.000	NaN	NaN	NaN	1992.392	768.556	50.000	1541.000	1947.000	2380.250	6924.000
DepartureDelayin_Mins	13572.000	NaN	NaN	NaN	18.076	46.841	0.000	0.000	0.000	14.000	1128.000
ArrivalDelayin_Mins	13508.000	NaN	NaN	NaN	18.384	46.923	0.000	0.000	0.000	15.000	1115.000
Satisfaction	13572.000	NaN	NaN	NaN	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Seat_comfort	13572.000	NaN	NaN	NaN	2.585	1.019	1.000	2.000	3.000	3.000	4.000
Departure.Arrival.time_convenient	12361.000	NaN	NaN	NaN	3.675	1.373	0.000	3.000	4.000	5.000	5.000
Food_drink	12352.000	NaN	NaN	NaN	2.559	1.159	0.000	2.000	3.000	3.000	5.000
Gate_location	13572.000	NaN	NaN	NaN	2.955	1.153	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	13572.000	NaN	NaN	NaN	3.038	1.378	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	13572.000	NaN	NaN	NaN	2.628	1.218	0.000	2.000	3.000	4.000	5.000
Online_support	13572.000	NaN	NaN	NaN	3.083	1.408	1.000	2.000	3.000	4.000	5.000
Ease_of_Onlinebooking	13572.000	NaN	NaN	NaN	3.061	1.385	1.000	2.000	3.000	4.000	5.000
Onboard_service	12521.000	NaN	NaN	NaN	3.309	1.271	1.000	2.000	3.000	4.000	5.000
Leg_room_service	13572.000	NaN	NaN	NaN	3.262	1.270	0.000	2.000	3.000	4.000	5.000
Baggage_handling	13572.000	NaN	NaN	NaN	3.686	1.163	1.000	3.000	4.000	5.000	5.000
Checkin_service	13572.000	NaN	NaN	NaN	3.342	1.258	1.000	3.000	3.000	4.000	5.000
Cleanliness	13572.000	NaN	NaN	NaN	3.709	1.149	1.000	3.000	4.000	5.000	5.000
Online_boarding	13572.000	NaN	NaN	NaN	3.087	1.383	1.000	2.000	3.000	4.000	5.000
Facilities_avg	13572.000	NaN	NaN	NaN	3.337	0.788	1.000	3.000	3.000	4.000	5.000
Online_service_avg	13572.000	NaN	NaN	NaN	3.078	1.290	1.000	2.000	3.000	4.000	5.000
InFlight_avg	13572.000	NaN	NaN	NaN	3.068	0.693	1.000	3.000	3.000	4.000	5.000

df_BT = df.query('TypeTravel == "Business travel" & Satisfaction==0')
df_BT.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	23490	2	Female	12396	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	20766	2	Loyal Customer	10794	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	22828.000	NaN	NaN	NaN	37.630	13.675	15.000	26.000	36.000	47.000	85.000
TypeTravel	23490	1	Business travel	23490	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	23490	3	Eco	11168	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	23490.000	NaN	NaN	NaN	2043.058	947.637	50.000	1498.000	1962.000	2509.000	6837.000
DepartureDelayin_Mins	23490.000	NaN	NaN	NaN	17.626	41.664	0.000	0.000	0.000	17.000	951.000
ArrivalDelayin_Mins	23430.000	NaN	NaN	NaN	18.505	42.247	0.000	0.000	1.000	18.000	952.000
Satisfaction	23490.000	NaN	NaN	NaN	0.000	0.000	0.000	0.000	0.000	0.000	0.000
Seat_comfort	23490.000	NaN	NaN	NaN	2.404	0.973	0.000	2.000	2.000	3.000	5.000
Departure.Arrival.time_convenient	21407.000	NaN	NaN	NaN	2.638	1.443	0.000	2.000	3.000	4.000	5.000
Food_drink	21410.000	NaN	NaN	NaN	2.727	1.280	0.000	2.000	3.000	4.000	5.000
Gate_location	23490.000	NaN	NaN	NaN	3.035	1.247	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	23490.000	NaN	NaN	NaN	2.854	1.328	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	23490.000	NaN	NaN	NaN	2.607	1.014	0.000	2.000	3.000	3.000	5.000
Online_support	23490.000	NaN	NaN	NaN	2.892	1.207	0.000	2.000	3.000	4.000	5.000
Ease_of_Onlinebooking	23490.000	NaN	NaN	NaN	2.737	1.245	0.000	2.000	3.000	4.000	5.000
Onboard_service	21568.000	NaN	NaN	NaN	2.784	1.221	0.000	2.000	3.000	4.000	5.000
Leg_room_service	23490.000	NaN	NaN	NaN	2.934	1.304	0.000	2.000	3.000	4.000	5.000
Baggage_handling	23490.000	NaN	NaN	NaN	3.187	1.085	1.000	3.000	3.000	4.000	5.000
Checkin_service	23490.000	NaN	NaN	NaN	2.759	1.241	0.000	2.000	3.000	4.000	5.000
Cleanliness	23490.000	NaN	NaN	NaN	3.193	1.088	0.000	3.000	3.000	4.000	5.000
Online_boarding	23490.000	NaN	NaN	NaN	2.747	1.271	0.000	2.000	3.000	4.000	5.000
Facilities_avg	23490.000	NaN	NaN	NaN	2.937	0.799	1.000	2.000	3.000	4.000	5.000
Online_service_avg	23490.000	NaN	NaN	NaN	2.793	1.125	1.000	2.000	3.000	3.000	5.000
InFlight_avg	23490.000	NaN	NaN	NaN	2.765	0.688	1.000	2.000	3.000	3.000	5.000

• For males, most of them loyal customers that travel for personal purposes, parameters like Seat comfort, food drink and in flight entertaiment are in between need improvement and acceptable level, that lead the to be neutral or unsatisfied with the service.
• For Females, most of them loyal customers that travel for business purposes, most of the parameters are in between need improvement and acceptable level, that lead the to be neutral or unsatisfied with the service.

Satisfaction vs Class

stacked_barplot(df, "Class", "Satisfaction")

Satisfaction      0      1    All
Class                            
All           41156  49761  90917
Eco           24755  16003  40758
Business      12600  30935  43535
Eco Plus       3801   2823   6624
------------------------------------------------------------------------------------------------------------------------

• Business Travelers are the most satisfied between the Class with a 71.05%

Satisfaccion vs Seat Comfort

stacked_barplot(df, "Seat_comfort", "Satisfaction")

Satisfaction      0      1    All
Seat_comfort                     
All           41156  49761  90917
3             13274   7278  20552
2             12904   7098  20002
1              7997   6690  14687
4              6878  12911  19789
5                97  12422  12519
0                 6   3362   3368
------------------------------------------------------------------------------------------------------------------------

• Good(4) and excellent (5) have a better percentage of satisfaction as would be expected.
• Most of the customers are in between acceptable (3) with 35.41% of satisfaction level and need improvement (2) with 35.48% of satisfaction level.
• not following a logic pattern, it is a positive satisfaction overall, for customers that think the Seat Comfort is poor (1) and extremely poor (0)

Satisfaction vs Departure Arrival Time convenient

stacked_barplot(df, "Departure.Arrival.time_convenient", "Satisfaction")

Satisfaction                           0      1    All
Departure.Arrival.time_convenient                     
All                                37501  45172  82673
4.0                                 8963   9877  18840
5.0                                 7552   9527  17079
3.0                                 6873   7933  14806
2.0                                 6789   7750  14539
1.0                                 5391   7819  13210
0.0                                 1933   2266   4199
------------------------------------------------------------------------------------------------------------------------

• In all the levels of satisfaction for Departure Arrival time convinient more than the ~53% of the customers are satisfied with the airline service overall.

Satisfaction vs Food Drink

stacked_barplot(df, "Food_drink", "Satisfaction")

Satisfaction      0      1    All
Food_drink                       
All           37469  45267  82736
3.0           10278   7713  17991
2.0            9903   7456  17359
4.0            7091  10154  17245
1.0            6521   6879  13400
5.0            2851  10096  12947
0.0             825   2969   3794
------------------------------------------------------------------------------------------------------------------------

• ~43% People that think the Food drink service is acceptable and need improvment are satisfied with the service overall.
• ~50% People that think the Food drink service is poor are satisfied with the service overall.
• ~60% People that think the Food drink service is good and ~80% that thinks is excellent are satisfied with the service overall.
• Despite the logic patterns, ~80% that qualified the food and drink as extremely poor are satisfied with the service.

Satisfaction vs Gate location

stacked_barplot(df, "Gate_location", "Satisfaction")

Satisfaction       0      1    All
Gate_location                     
All            41156  49761  90917
3              12585  10800  23385
4              10621  10467  21088
2               7222   9891  17113
1               6133   9743  15876
5               4595   8859  13454
0                  0      1      1
------------------------------------------------------------------------------------------------------------------------

• ~55% of the travelers that qualified the gate location as manageable and convinient said that they have a positive satisfaction with the service overall.
• ~60% of the travelers that qualified the gate location as need improvement and very inconvenient, showed a positive satisfaction overall.
• ~70% of the travelers that thinks the gate location in very convinient, showed a positive satisfaction overall.

Satisfaction vs In flight wifi

stacked_barplot(df, "Inflightwifi_service", "Satisfaction")

Satisfaction              0      1    All
Inflightwifi_service                     
All                   41156  49761  90917
2                      9447   9447  18894
3                      9382   9817  19199
4                      7994  14165  22159
1                      7527   2784  10311
5                      6753  13505  20258
0                        53     43     96
------------------------------------------------------------------------------------------------------------------------

• ~30% of the travellers that thinks the WiFi service is poor have a positive satisfaction overall.
• The satisfaction overall increases as long as the In flight WI Fi satisfaction improves.

Satisfaction vs Inflight Entertainment

stacked_barplot(df, "Inflight_entertainment", "Satisfaction")

Satisfaction                0      1    All
Inflight_entertainment                     
All                     41156  49761  90917
3                       13641   3354  16995
2                       11181   2346  13527
4                        8186  21187  29373
1                        6459   1739   8198
5                        1011  19775  20786
0                         678   1360   2038
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the In flight Etertaiment satisfaction improves.
• Despite the logic, travellers that qualified the Inflight enterteiment as extremely poor, still are satisfied in a ~65% overall.

Satisfaction vs Online Support

stacked_barplot(df, "Online_support", "Satisfaction")

Satisfaction        0      1    All
Online_support                     
All             41156  49761  90917
3               10834   4256  15090
4                9229  19813  29042
2                8501   3562  12063
1                6867   2938   9805
5                5724  19192  24916
0                   1      0      1
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the Online Support satisfaction improves.
• For level extremely poor, poor and acceptable, the satisfaction proportion is constant in a ~ 30%.

Satisfaction vs Ease Online booking

stacked_barplot(df, "Ease_of_Onlinebooking", "Satisfaction")

Satisfaction               0      1    All
Ease_of_Onlinebooking                     
All                    41156  49761  90917
3                      10067   5619  15686
2                       9944   3952  13896
4                       7867  20126  27993
1                       7560   1810   9370
5                       5706  18254  23960
0                         12      0     12
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the Ease Online booking satisfaction improves.

Satisfaction vs Onboard Service

stacked_barplot(df, "Onboard_service", "Satisfaction")

Satisfaction         0      1    All
Onboard_service                     
All              37910  45828  83738
3.0              10278   7133  17411
4.0               9353  17020  26373
2.0               7271   3747  11018
1.0               6236   2301   8537
5.0               4769  15627  20396
0.0                  3      0      3
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the onboard service satisfaction improves.

Satisfaction vs Leg room service

stacked_barplot(df, "Leg_room_service", "Satisfaction")

Satisfaction          0      1    All
Leg_room_service                     
All               41156  49761  90917
3                  9952   5823  15775
2                  9475   5681  15156
4                  9051  18763  27814
5                  7018  17053  24071
1                  5556   2223   7779
0                   104    218    322
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the leg room service satisfaction improves.
• Despite the logic pattern, traveller that qualified the leg room service level as poor, have a ~70% satisfaction level

Satisfaction vs Baggage Handling

stacked_barplot(df, "Baggage_handling", "Satisfaction")

Satisfaction          0      1    All
Baggage_handling                     
All               41156  49761  90917
4                 13927  19895  33822
3                 11771   5462  17233
5                  6629  18373  25002
2                  5627   3674   9301
1                  3202   2357   5559
------------------------------------------------------------------------------------------------------------------------

• The travellers that qualified the handling baggage as excellent have the highest satisfaction level proportion at ~75%.
• The travellers that qualified the handling baggage as acceptable have the lowest satisfaction level proportion at ~35%.

Satisfaction vs Check in service

stacked_barplot(df, "Checkin_service", "Satisfaction")

Satisfaction         0      1    All
Checkin_service                     
All              41156  49761  90917
3                10832  14109  24941
4                10728  14755  25483
1                 7352   3409  10761
2                 7238   3575  10813
5                 5005  13913  18918
0                    1      0      1
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the check in service satisfaction improves

Satisfaction vs Cleanliness

stacked_barplot(df, "Cleanliness", "Satisfaction")

Satisfaction      0      1    All
Cleanliness                      
All           41156  49761  90917
4             14151  20095  34246
3             11543   5387  16930
5              6730  18349  25079
2              5544   3739   9283
1              3184   2191   5375
0                 4      0      4
------------------------------------------------------------------------------------------------------------------------

• The travellers that qualified the Cleanliness as excellent have the highest satisfaction level proportion at ~75%.
• The travellers that qualified the Cleanliness as acceptable have the lowest satisfaction level proportion at ~35%, no considering the ones that qualified the Cleanliness extremely poor that it was only 4 travellers no representative enough number for the data.

Online Boarding vs Satisfaction

stacked_barplot(df, "Online_boarding", "Satisfaction")

Satisfaction         0      1    All
Online_boarding                     
All              41156  49761  90917
3                 9636  11791  21427
2                 9356   3679  13035
4                 8516  16160  24676
1                 7904   2873  10777
5                 5735  15258  20993
0                    9      0      9
------------------------------------------------------------------------------------------------------------------------

• The satisfaction overall increases as long as the Online Boarding satisfaction improves

Survey Variables Summary

Facilities Average vs Satisfaction

stacked_barplot(df, "Facilities_avg", "Satisfaction")

Satisfaction        0      1    All
Facilities_avg                     
All             41156  49761  90917
3.0             15738  12307  28045
4.0             13840  30249  44089
2.0             10774   4516  15290
5.0               491   2608   3099
1.0               313     81    394
------------------------------------------------------------------------------------------------------------------------

• Passangers that qualified the Facilities on average between extremely poor and poor are most likely disatisfied.
• Passangers that qualified the Facilities as good and excellent are most likely satisfied.

sns.set(font_scale=1)
fig, axarr = plt.subplots(2, 2, figsize=(12, 12))

parameters = {"axes.labelsize": 15, "axes.titlesize": 10}
plt.rcParams.update(parameters)

table1 = pd.crosstab(df["Satisfaction"], df["Gate_location"])
sns.heatmap(table1, cmap="Oranges", ax=axarr[0][0])
table2 = pd.crosstab(df["Satisfaction"], df["Onboard_service"])
sns.heatmap(table2, cmap="Blues", ax=axarr[0][1])
table3 = pd.crosstab(df["Satisfaction"], df["Baggage_handling"])
sns.heatmap(table3, cmap="pink", ax=axarr[1][0])
table4 = pd.crosstab(df["Satisfaction"], df["Checkin_service"])
sns.heatmap(table4, cmap="bone", ax=axarr[1][1])
# table5 = pd.crosstab(df["Satisfaction"], df["Facilities_avg"])
# sns.heatmap(table5, cmap="red", ax=axarr[2][0])

<AxesSubplot:xlabel='Checkin_service', ylabel='Satisfaction'>

table5 = pd.crosstab(df["Satisfaction"], df["Facilities_avg"])
sns.heatmap(table5, cmap="Oranges")

<AxesSubplot:xlabel='Facilities_avg', ylabel='Satisfaction'>

• For Gate Location, passangers are predominantly satisfied, but Theres a high population that thinks the gate location is convinient.
• For Onboard service, bagging handling and Checking service, people are dissatisfied from poor to acceptable. Passanger that think is good to excelent are satisfied with the service.
• Overall, passangers that rated the service between needs improvement to acceptable are predominantly dissatisfied with the service, while people that think is good are most likely satisfied with the service

Online Service Average vs Satisfaction

stacked_barplot(df, "Online_service_avg", "Satisfaction")

Satisfaction            0      1    All
Online_service_avg                     
All                 41156  49761  90917
3.0                 12498   7445  19943
2.0                 10421   3059  13480
4.0                  8057  25820  33877
1.0                  5544   1155   6699
5.0                  4636  12282  16918
------------------------------------------------------------------------------------------------------------------------

• Overall customer thinks the online service is good to excellent are satisfied with the service.

fig, axarr = plt.subplots(2, 2, figsize=(12, 12))
table6 = pd.crosstab(df["Satisfaction"], df["Online_support"])
sns.heatmap(table6, cmap="Oranges", ax=axarr[0][0])
table7 = pd.crosstab(df["Satisfaction"], df["Ease_of_Onlinebooking"])
sns.heatmap(table7, cmap="Blues", ax=axarr[0][1])
table8 = pd.crosstab(df["Satisfaction"], df["Online_boarding"])
sns.heatmap(table8, cmap="pink", ax=axarr[1][0])
table9 = pd.crosstab(df["Satisfaction"], df["Online_service_avg"])
sns.heatmap(table9, cmap="bone", ax=axarr[1][1])

<AxesSubplot:xlabel='Online_service_avg', ylabel='Satisfaction'>

• For Online Support and Ease of Online Booking, passangers that think the service is between poor to acceptable most likely are dissatisfied, the ones that rated good to excellent are satisfied with the service.
• For Online boarding, passanger that think is poor to needs improvement are dissatisfied with the service.
• Overall, passengers that rated the service online between poor to acceptable are predominantly dissatisfied with the service, while people that think is good and excellent are most likely satisfied with the service.

Inflight Service Average vs Satisfaction

stacked_barplot(df, "InFlight_avg", "Satisfaction")

Satisfaction      0      1    All
InFlight_avg                     
All           41156  49761  90917
3.0           21628  17966  39594
2.0           11644   3890  15534
4.0            7370  23812  31182
1.0             467    586   1053
5.0              47   3504   3551
0.0               0      3      3
------------------------------------------------------------------------------------------------------------------------

• Overall, people that thinks the Inflight serbvice is good most likely are satisfied. However most of the passangers rated as acceptable are dissatisfied with the inflight service in general.

fig, axarr = plt.subplots(2, 2, figsize=(12, 12))

table10 = pd.crosstab(df['Satisfaction'], df['Seat_comfort'])
sns.heatmap(table10, cmap='Oranges', ax = axarr[0][0])
table11 = pd.crosstab(df['Satisfaction'], df['Departure.Arrival.time_convenient'])
sns.heatmap(table11, cmap='Blues', ax = axarr[0][1])
table12 = pd.crosstab(df['Satisfaction'], df['Food_drink'])
sns.heatmap(table12, cmap='pink', ax = axarr[1][0])
table13 = pd.crosstab(df['Satisfaction'], df['Inflightwifi_service'])
sns.heatmap(table13, cmap='bone', ax = axarr[1][1])

<AxesSubplot:xlabel='Inflightwifi_service', ylabel='Satisfaction'>

fig, axarr = plt.subplots(2, 2, figsize=(12, 12))
table14 = pd.crosstab(df["Satisfaction"], df["Inflight_entertainment"])
sns.heatmap(table14, cmap="Oranges", ax=axarr[0][0])
table15 = pd.crosstab(df["Satisfaction"], df["Leg_room_service"])
sns.heatmap(table15, cmap="Blues", ax=axarr[0][1])
table16 = pd.crosstab(df["Satisfaction"], df["Cleanliness"])
sns.heatmap(table16, cmap="pink", ax=axarr[1][0])
table17 = pd.crosstab(df["Satisfaction"], df["InFlight_avg"])
sns.heatmap(table17, cmap="bone", ax=axarr[1][1])

<AxesSubplot:xlabel='InFlight_avg', ylabel='Satisfaction'>

• For Seat Comfort, Inflight Entertaiment and Ease of Online Booking, passangers that think the service is between poor to acceptable most likely are dissatisfied, the ones that rated good to excellent are satisfied with the service.
• Overall, passangers that rated the Inflight between poor to acceptable are predominantly dissatisfied with the service, while people that think is good and excellent are most likely satisfied with the service.

Analysis of Bivariate numerical variables

### Function to plot distributions


def distribution_plot_wrt_target(df, predictor, target):

    fig, axs = plt.subplots(2, 2, figsize=(12, 10))

    target_uniq = df[target].unique()

    axs[0, 0].set_title("Distribution of target for target=" + str(target_uniq[0]))
    sns.histplot(
        data=df[df[target] == target_uniq[0]],
        x=predictor,
        kde=True,
        ax=axs[0, 0],
        color="teal",
    )

    axs[0, 1].set_title("Distribution of target for target=" + str(target_uniq[1]))
    sns.histplot(
        data=df[df[target] == target_uniq[1]],
        x=predictor,
        kde=True,
        ax=axs[0, 1],
        color="orange",
    )

    axs[1, 0].set_title("Boxplot w.r.t target")
    sns.boxplot(data=df, x=target, y=predictor, ax=axs[1, 0], palette="gist_rainbow")

    axs[1, 1].set_title("Boxplot (without outliers) w.r.t target")
    sns.boxplot(
        data=df,
        x=target,
        y=predictor,
        ax=axs[1, 1],
        showfliers=False,
        palette="gist_rainbow",
    )

    plt.tight_layout()
    plt.show()

Satisfaction vs Age

distribution_plot_wrt_target(df, "Age", "Satisfaction")

Histogram more in detail

sns.set(font_scale=1)
with sns.axes_style("white"):
    g = sns.catplot(
        "Age",
        data=df,
        aspect=3.0,
        kind="count",
        hue="Satisfaction",
        order=range(15, 80),
    )
    g.set_ylabels("Age vs Passenger Satisfaction")

• From the histogram distribution for the travellers with positive satisfaction overall is skwed to the left, meaning that most of the people satisfied are between ~39 to 60. Same logic for the boxplot, where the average is ~43 years old.
• From the histogram distribution for the travellers with neutral or dissatisfied overall is skwed to the right, meaning that most of the people neutral or dissatisfied are younger 40. Same logic for the boxplot, where the average is ~35 years old.
• For passangers between 61 to 79, passangers are most likely dissatisfied

Let's check what make them satisfied or dissatisfied

Age_young = df[df.Age < 39]
Age_middle = df[(df.Age > 38) | (df.Age < 61)]
Age_old = df[df.Age > 60]

Age_young.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	37730	2	Female	19370	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	33996	2	Loyal Customer	22565	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	37730.000	NaN	NaN	NaN	27.748	6.404	15.000	23.000	27.000	33.000	38.000
TypeTravel	33969	2	Business travel	24433	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	37730	3	Eco	18877	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	37730.000	NaN	NaN	NaN	2258.925	931.695	50.000	1659.000	2072.000	2664.000	6907.000
DepartureDelayin_Mins	37730.000	NaN	NaN	NaN	14.825	38.751	0.000	0.000	0.000	12.000	1305.000
ArrivalDelayin_Mins	37619.000	NaN	NaN	NaN	15.302	39.349	0.000	0.000	0.000	13.000	1280.000
Satisfaction	37730.000	NaN	NaN	NaN	0.462	0.499	0.000	0.000	0.000	1.000	1.000
Seat_comfort	37730.000	NaN	NaN	NaN	2.806	1.382	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	34279.000	NaN	NaN	NaN	2.877	1.572	0.000	2.000	3.000	4.000	5.000
Food_drink	34335.000	NaN	NaN	NaN	2.813	1.431	0.000	2.000	3.000	4.000	5.000
Gate_location	37730.000	NaN	NaN	NaN	2.992	1.271	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	37730.000	NaN	NaN	NaN	3.231	1.347	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	37730.000	NaN	NaN	NaN	3.148	1.397	0.000	2.000	3.000	4.000	5.000
Online_support	37730.000	NaN	NaN	NaN	3.317	1.359	1.000	2.000	4.000	4.000	5.000
Ease_of_Onlinebooking	37730.000	NaN	NaN	NaN	3.342	1.341	0.000	2.000	4.000	4.000	5.000
Onboard_service	34759.000	NaN	NaN	NaN	3.331	1.286	1.000	2.000	4.000	4.000	5.000
Leg_room_service	37730.000	NaN	NaN	NaN	3.317	1.319	0.000	2.000	3.000	4.000	5.000
Baggage_handling	37730.000	NaN	NaN	NaN	3.676	1.132	1.000	3.000	4.000	5.000	5.000
Checkin_service	37730.000	NaN	NaN	NaN	3.265	1.283	1.000	2.000	3.000	4.000	5.000
Cleanliness	37730.000	NaN	NaN	NaN	3.689	1.134	1.000	3.000	4.000	5.000	5.000
Online_boarding	37730.000	NaN	NaN	NaN	3.279	1.339	0.000	2.000	3.000	4.000	5.000
Facilities_avg	37730.000	NaN	NaN	NaN	3.323	0.802	1.000	3.000	3.000	4.000	5.000
Online_service_avg	37730.000	NaN	NaN	NaN	3.314	1.258	1.000	2.000	4.000	4.000	5.000
InFlight_avg	37730.000	NaN	NaN	NaN	3.136	0.825	0.000	3.000	3.000	4.000	5.000

Age_middle.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	86429	2	Female	43910	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	77775	2	Loyal Customer	63323	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	86429.000	NaN	NaN	NaN	40.945	13.967	15.000	29.000	41.000	52.000	85.000
TypeTravel	77819	2	Business travel	55585	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	86429	3	Business	42884	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	86429.000	NaN	NaN	NaN	1968.643	1036.967	50.000	1325.000	1915.000	2541.000	6950.000
DepartureDelayin_Mins	86429.000	NaN	NaN	NaN	14.653	38.502	0.000	0.000	0.000	12.000	1592.000
ArrivalDelayin_Mins	86164.000	NaN	NaN	NaN	15.020	38.881	0.000	0.000	0.000	13.000	1584.000
Satisfaction	86429.000	NaN	NaN	NaN	0.554	0.497	0.000	0.000	1.000	1.000	1.000
Seat_comfort	86429.000	NaN	NaN	NaN	2.840	1.395	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	78596.000	NaN	NaN	NaN	2.981	1.525	0.000	2.000	3.000	4.000	5.000
Food_drink	78680.000	NaN	NaN	NaN	2.854	1.443	0.000	2.000	3.000	4.000	5.000
Gate_location	86429.000	NaN	NaN	NaN	2.991	1.312	0.000	2.000	3.000	4.000	5.000
Inflightwifi_service	86429.000	NaN	NaN	NaN	3.257	1.319	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	86429.000	NaN	NaN	NaN	3.400	1.336	0.000	2.000	4.000	4.000	5.000
Online_support	86429.000	NaN	NaN	NaN	3.533	1.302	0.000	3.000	4.000	5.000	5.000
Ease_of_Onlinebooking	86429.000	NaN	NaN	NaN	3.485	1.301	0.000	2.000	4.000	5.000	5.000
Onboard_service	79557.000	NaN	NaN	NaN	3.471	1.270	0.000	3.000	4.000	4.000	5.000
Leg_room_service	86429.000	NaN	NaN	NaN	3.493	1.290	0.000	2.000	4.000	5.000	5.000
Baggage_handling	86429.000	NaN	NaN	NaN	3.695	1.158	1.000	3.000	4.000	5.000	5.000
Checkin_service	86429.000	NaN	NaN	NaN	3.342	1.260	0.000	3.000	3.000	4.000	5.000
Cleanliness	86429.000	NaN	NaN	NaN	3.704	1.152	0.000	3.000	4.000	5.000	5.000
Online_boarding	86429.000	NaN	NaN	NaN	3.360	1.297	0.000	2.000	4.000	4.000	5.000
Facilities_avg	86429.000	NaN	NaN	NaN	3.377	0.817	1.000	3.000	4.000	4.000	5.000
Online_service_avg	86429.000	NaN	NaN	NaN	3.459	1.159	1.000	3.000	4.000	4.000	5.000
InFlight_avg	86429.000	NaN	NaN	NaN	3.230	0.816	0.000	3.000	3.000	4.000	5.000

Age_old.describe(include="all").T

	count	unique	top	freq	mean	std	min	25%	50%	75%	max
Gender	7128	2	Female	3581	NaN	NaN	NaN	NaN	NaN	NaN	NaN
CustomerType	6406	2	Loyal Customer	6084	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Age	7128.000	NaN	NaN	NaN	66.105	4.148	61.000	63.000	66.000	69.000	85.000
TypeTravel	6420	2	Personal Travel	3982	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Class	7128	3	Eco	4468	NaN	NaN	NaN	NaN	NaN	NaN	NaN
Flight_Distance	7128.000	NaN	NaN	NaN	1525.207	1003.817	50.000	608.000	1511.000	2171.000	6889.000
DepartureDelayin_Mins	7128.000	NaN	NaN	NaN	14.177	36.030	0.000	0.000	0.000	12.000	565.000
ArrivalDelayin_Mins	7099.000	NaN	NaN	NaN	14.502	36.582	0.000	0.000	0.000	12.000	624.000
Satisfaction	7128.000	NaN	NaN	NaN	0.433	0.496	0.000	0.000	0.000	1.000	1.000
Seat_comfort	7128.000	NaN	NaN	NaN	2.786	1.358	0.000	2.000	3.000	4.000	5.000
Departure.Arrival.time_convenient	6486.000	NaN	NaN	NaN	3.197	1.487	0.000	2.000	3.000	4.000	5.000
Food_drink	6464.000	NaN	NaN	NaN	2.815	1.448	0.000	2.000	3.000	4.000	5.000
Gate_location	7128.000	NaN	NaN	NaN	2.964	1.289	1.000	2.000	3.000	4.000	5.000
Inflightwifi_service	7128.000	NaN	NaN	NaN	3.141	1.341	0.000	2.000	3.000	4.000	5.000
Inflight_entertainment	7128.000	NaN	NaN	NaN	3.227	1.313	0.000	2.000	3.000	4.000	5.000
Online_support	7128.000	NaN	NaN	NaN	3.413	1.292	1.000	2.000	4.000	4.000	5.000
Ease_of_Onlinebooking	7128.000	NaN	NaN	NaN	3.341	1.326	1.000	2.000	4.000	4.000	5.000
Onboard_service	6559.000	NaN	NaN	NaN	3.357	1.296	1.000	2.000	4.000	4.000	5.000
Leg_room_service	7128.000	NaN	NaN	NaN	3.410	1.306	0.000	2.000	4.000	4.000	5.000
Baggage_handling	7128.000	NaN	NaN	NaN	3.552	1.225	1.000	3.000	4.000	5.000	5.000
Checkin_service	7128.000	NaN	NaN	NaN	3.230	1.276	1.000	2.000	3.000	4.000	5.000
Cleanliness	7128.000	NaN	NaN	NaN	3.562	1.197	1.000	3.000	4.000	4.000	5.000
Online_boarding	7128.000	NaN	NaN	NaN	3.196	1.320	1.000	2.000	3.000	4.000	5.000
Facilities_avg	7128.000	NaN	NaN	NaN	3.276	0.839	1.000	3.000	3.000	4.000	5.000
Online_service_avg	7128.000	NaN	NaN	NaN	3.315	1.127	1.000	3.000	3.000	4.000	5.000
InFlight_avg	7128.000	NaN	NaN	NaN	3.169	0.797	1.000	3.000	3.000	4.000	5.000

• Young and Older Age: mostly are dissatisfied mostly for the Inflight service and Facilities, They travel in ECO Class most likely.
• Middle Age: mostly travel in Business class.

Satisfaction vs Flight distance

distribution_plot_wrt_target(df, "Flight_Distance", "Satisfaction")

• The histogram of flight distance for satisfied travellers seem to be bimodal and from boxplot outliers start at ~4000.
• The histogram of flight distance for neutral or dissatisfied travellers seem to be skwed to the right and from boxplot outliers start at ~5000.
• From the boxplot both, satisfied and neutral or dissatisfied, have the same average, however the distance between Q2 and Q3 is wider for the satisfied data.

Satisfaction vs Departure Delay in Mins

distribution_plot_wrt_target(df, "DepartureDelayin_Mins", "Satisfaction")

• There's not much difference from the histogram for both
• From the boxplot, there's not much of differents for both, however for neutral or dissatisfied the data range is wider as well as the outliers.

Satisfaction vs Arrival Delay in Mins¶

distribution_plot_wrt_target(df, "ArrivalDelayin_Mins", "Satisfaction")

• There's not much difference from the histogram for both
• From the boxplot, there's not much of differents for both, however for neutral or dissatisfied the data range is wider as well as the outliers.

Multivariate Analysis

Customers Information

cols = df[
    ["Age", "Flight_Distance", "DepartureDelayin_Mins", "ArrivalDelayin_Mins"]
].columns.tolist()
plt.figure(figsize=(12, 10))
for i, variable in enumerate(cols):
    plt.subplot(3, 3, i + 1)
    sns.lineplot(df["Age"], df[variable], hue=df["Satisfaction"], ci=0)
    plt.tight_layout()
    plt.title(variable)
plt.show()

Flight distance vs Age vs Satisfaction

• Younger people (under 40) regardless their satisfaction level tend to travel longer distance (over ~2000) and their level of satisfaction is higher proportional with the distance, in general.
• Older people (above 40) regardless their satisfaction level tend to travel shorter distance (below ~2000) and their level of satisfaction is less proportional with the distance, in general.

Departure Delay in Mins vs Age vs Satisfaction

• Departure delay in Mins has a neutral or dissatisfied level of satisfaction level is showed when the departure delay is longer and normaly above 15 mins, regardless of the age, as should be expected.

Arrival Delay in Mins vs Age vs Satisfaction

• Arrival delay in Mins has a neutral or dissatisfied level of satisfaction level is showed when the arrival delay is longer and normally above 15 min, regardless of the age, as should be expected.

Type Travel vs Age vs Gender vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Age",
        y="TypeTravel",
        hue="Satisfaction",
        col="Gender",
        data=df,
        kind="bar",
        height=5,
        aspect=1,
    )

• Type Travel, Class, Gender: regardless the gender, Passangers that travel for business are more satisfied with the service for older agers. For personal purpose remains equal.

Customer Type vs Age vs Gender vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Age",
        y="CustomerType",
        hue="Satisfaction",
        col="Gender",
        data=df,
        kind="bar",
        height=5,
        aspect=1,
    )

• Loyal Customer tend to be older, regardless the gender and are equally satisfied or dissatisfied with the service.
• Disloyal customers are under 30 on average and most likely dissatisfied.

Customer Type vs Flight Distance vs Gender vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Flight_Distance",
        y="CustomerType",
        hue="Satisfaction",
        col="Gender",
        data=df,
        kind="bar",
        height=5,
        aspect=1,
    )

• For Flight Distance Loyal and disloyal Passangers are equally satisfied or dissatisfied.
• Loyal Female Passanger travels on average less distance than disloyals.

Type Travel vs Class vs Age vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Age",
        y="TypeTravel",
        hue="Satisfaction",
        col="Class",
        data=df,
        kind="bar",
        height=5,
        aspect=1,
    )

• Type Travel, Class, Age: For Business travellers, regarless the class the satisfaction is slighly highier than the disatisfaction in older ages. For other combinations is almost equal the satisfaction level.

Customer Type vs Class vs Age vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Age",
        y="CustomerType",
        hue="Satisfaction",
        col="Class",
        data=df,
        kind="bar",
        height=5,
        aspect=1,
    )

• The neutral or dissatisfaction is higher in older ages for disloyal customers regardless the age but quite significant for business class customers

Flight distance vs Type of Travel vs Class vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Flight_Distance",
        y="TypeTravel",
        hue="Satisfaction",
        col="Class",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• Type Travel, Class, Flight Distance. Business and Personal passangers are more dissatisfied with Eco and Eco Plus for longer distance. For the other combinations, the distribution between satisffied and dissatisfied is almost equal.

Flight distance vs Customer Type vs Class vs Satisfaction¶

with sns.axes_style("white"):
    g = sns.catplot(
        x="Flight_Distance",
        y="CustomerType",
        hue="Satisfaction",
        col="Class",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• Disloyal Passangers are equally satisfied and dissatisfied, travel around 2000 miles on average .
• For Loyal Customers the satisfied ones travel less miles than the dissatisfied ones.

Departure Delay in Mins vs Type of Travel vs Class vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Class",
        y="DepartureDelayin_Mins",
        hue="Satisfaction",
        col="TypeTravel",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• For business passangers have more than ~12 mins of delay, is a great signal for dissatisfaction, regardless the class.
• For personal passangers have more than ~10 mins of delay, is a great signal for dissatisfaction, regardless the class.

Arrival Delay in Mins vs Type of Travel vs Class vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Class",
        y="ArrivalDelayin_Mins",
        hue="Satisfaction",
        col="TypeTravel",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• For business passangers have more than ~12 mins of delay, is a great signal for dissatisfaction, regardless the class.-
• For personal passangers have more than ~10 mins of delay, is a great signal for dissatisfaction, regardless the class.

Age vs Facilities vs Satisfaction

cols = df[
    [
        "Gate_location",
        "Onboard_service",
        "Baggage_handling",
        "Checkin_service",
        "Facilities_avg",
    ]
].columns.tolist()
plt.figure(figsize=(12, 10))
for i, variable in enumerate(cols):
    plt.subplot(3, 3, i + 1)
    sns.lineplot(df["Age"], df[variable], hue=df["Satisfaction"], ci=0)
    plt.tight_layout()
    plt.title(variable)
plt.show()

Gate Location vs Age vs Satisfaction

• In general, Gate location is qualified as manageable, however for older people (over 70) this could be qualified as need improvement, but still be qualified as positive satisfaction in most of the cases.

Onboard Service vs Age vs Satisfaction

• In general, Satisfied travellers qualified onboard service between acceptable to good regardless the age.
• Neutral and dissatisfied travellers qualified onboard service between acceptable to needs improvement regardless the age.

Baggage Handling vs Age vs Satisfaction

• In general, Satisfied travellers qualified baggage handling as good regardless the age.
• In general, Neutral and dissatisfied travellers qualified baggage handling as needs improvement to good regardless the age.

Checking Service vs Age vs Satisfaction

• In general, Satisfied travellers qualified check in service between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travellers qualified check in service between acceptable to needs improvement regardless the age.

Facilities Average vs Age vs Satisfaction

• In general, Satisfied travellers qualified Facilities service between acceptable to good regardless the age.
• Neutral and dissatisfied travellers qualified Facilities service between acceptable to needs improvement regardless the age. The drop on ages around 80 is substantial probably derivated for the poor rate and dissatisfaction specially for Onboard services, baggage handling and Checking service

Age vs Online Service vs Satisfaction

cols = df[
    ["Online_support", "Ease_of_Onlinebooking", "Online_boarding", "Online_service_avg"]
].columns.tolist()
plt.figure(figsize=(12, 10))
for i, variable in enumerate(cols):
    plt.subplot(3, 3, i + 1)
    sns.lineplot(df["Age"], df[variable], hue=df["Satisfaction"], ci=0)
    plt.tight_layout()
    plt.title(variable)
plt.show()

Online Support vs Age vs Satisfaction

• In general, Satisfied travelers qualified online support between acceptable to excellent regardless the age.
• In general, Neutral and dissatisfied travelers qualified online service between acceptable to needs improvement regardless the age.

Ease of Online Booking vs Age vs Satisfaction

• In general, Satisfied travelers qualified Ease of Online Booking between acceptable to excellent regardless the age.
• In general, Neutral and dissatisfied travelers qualified Ease of Online Booking between acceptable to needs improvement regardless the age.

Online boarding vs Age vs Satisfaction

• In general, Satisfied travelers qualified Online boarding between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travelers qualified Online boarding between acceptable to needs improvement regardless the age.

Online Service Average vs Age vs Satisfaction

• In general, Satisfied travelers qualified Online service average as good regardless the age.
• In general, Neutral and dissatisfied travelers qualified Online service average between acceptable to needs improvement regardless the age.

Age vs In-Flight vs Satisfaction

cols = df[
    [
        "Seat_comfort",
        "Departure.Arrival.time_convenient",
        "Food_drink",
        "Inflightwifi_service",
        "Inflight_entertainment",
        "Leg_room_service",
        "Cleanliness",
        "InFlight_avg",
    ]
].columns.tolist()
plt.figure(figsize=(12, 10))
for i, variable in enumerate(cols):
    plt.subplot(3, 3, i + 1)
    sns.lineplot(df["Age"], df[variable], hue=df["Satisfaction"], ci=0)
    plt.tight_layout()
    plt.title(variable)
plt.show()

Seat comfort vs Age vs Satisfaction

• In general, Satisfied travelers qualified Seat comfort between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travelers qualified Seat comfort between acceptable to needs improvement regardless the age.

Departure Arrival Time convenien vs Age vs Satisfaction

• In general, there's not much difference between satisfied or neutral /dissatisfied, which is between needs improvement to good.

Food Drink vs Age vs Satisfaction

• In general, Satisfied travelers food and drink between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travelers qualified food and drink between acceptable to needs improvement regardless the age.

Inflight Wi-Fi service vs Age vs Satisfaction

• In general, Satisfied travelers qualified Inflight Wi-Fi service between acceptable to excellent regardless the age.
• In general, Neutral and dissatisfied travelers qualified Inflight Wi-Fi service between acceptable to needs improvement regardless the age.

Inflight Entertainment vs Age vs Satisfaction

• In general, Satisfied travelers qualified Inflight Entertainment between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travelers qualified Inflight Entertainment between acceptable to needs improvement regardless the age.

Leg Room vs Age vs Satisfaction

• In general, Satisfied travelers qualified Leg room between acceptable to excellent.
• In general, Neutral and dissatisfied travelers qualified Leg room between acceptable to needs improvement regardless the age.

Cleanliness vs Age vs Satisfaction

• In general, Satisfied travelers qualified Cleanliness between good to excellent regardless the age.
• In general, Neutral and dissatisfied travelers qualified Cleanliness between good to acceptable regardless the age.

Inflight Average vs Age vs Satisfaction

• In general, Satisfied travelers qualified Inflight Average between acceptable to good regardless the age.
• In general, Neutral and dissatisfied travelers qualified Inflight Average between good to poor to acceptable regardless the age.
• Seat comfort Departure/Arriveal time convinient, food and drink and Inflight WiFi weight a lot on the customer dissatisfaction.

Inflight Entertainment vs Flight Distance vs Type Travel vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Inflight_entertainment",
        y="Flight_Distance",
        hue="Satisfaction",
        col="TypeTravel",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• IE [Poor-Good], for Personal purposes travel that travel longer distance(A1800 miles/3.5hrs) are most likely dissatisfied. While, for Business purposes this trend happens for passengers that rated IA as Extremely poor to needs improvement.

Inflight Entertainment vs Age vs Gender vs Satisfaction

with sns.axes_style("white"):
    g = sns.catplot(
        x="Inflight_entertainment",
        y="Age",
        hue="Satisfaction",
        col="Gender",
        data=df,
        kind="bar",
        height=4.5,
        aspect=0.8,
    )

• IE [Poor-Acceptable] Younger Females (U40) most likely are dissatisfied with the service. Younger Males that rated as [Poor-Need Improvement] are dissatisfied.

Conclusions

• The data is skewed towards the Business class customers insights, since most likely, in an airplane there are more people flying in Eco or Eco plus class than in business class.
• In flight Entertainment, Check in service, Online boarding, Onboard service are the parameters that stands out for Business class customer be satisfied with the airline services than for customers in Eco or Eco plus class.
• There’s loyal customer that are neutral or unsatisfied with the airline service, since most of them are males traveling for personal reasons in Eco class, the company should target to offer them discounts or perks in the place of arrival. This cluster is a group on risk to look for another preferred airline.
• Focus on disloyal customers that are satisfied with the airline service. The company should create an incentive program to become this target as potential loyal customer, meaning a returning customer.
• Younger people (under 40) tend to be satisfied with longer distance flights, while older people tend to be satisfied with shorter distance flights.
• Departure/ Arrival delay in Mins has a neutral or dissatisfied level of satisfaction level is showed when the departure/arrival delay is longer and normally above 15, regardless of the age, as should be expected. Airline should work in minimize this waiting time to be under 15 mins.
• Airline should be more proactive to reach out people over 70 years old and guide them in regards of gate location, baggage handling, on board services and check-in service.

Let's find the percentage of outliers, in each column of the data, using IQR.

Q1 = df.quantile(0.25)  # To find the 25th percentile and 75th percentile.
Q3 = df.quantile(0.75)

IQR = Q3 - Q1  # Inter Quantile Range (75th perentile - 25th percentile)

lower = (
    Q1 - 1.5 * IQR
)  # Finding lower and upper bounds for all values. All values outside these bounds are outliers
upper = Q3 + 1.5 * IQR

(
    (df.select_dtypes(include=["float64", "int64"]) < lower)
    | (df.select_dtypes(include=["float64", "int64"]) > upper)
).sum() / len(df) * 100

Age                                  0.000
Flight_Distance                      1.980
DepartureDelayin_Mins               13.881
ArrivalDelayin_Mins                 13.403
Satisfaction                         0.000
Seat_comfort                         0.000
Departure.Arrival.time_convenient    0.000
Food_drink                           0.000
Gate_location                        0.000
Inflightwifi_service                 0.000
Inflight_entertainment               0.000
Online_support                       0.000
Ease_of_Onlinebooking                0.000
Onboard_service                      9.393
Leg_room_service                     0.000
Baggage_handling                     0.000
Checkin_service                     11.837
Cleanliness                          0.000
Online_boarding                      0.000
Facilities_avg                       0.433
Online_service_avg                   7.368
InFlight_avg                         1.161
dtype: float64

• After identifying outliers, we can decide whether to remove/treat them or not. It depends on one's approach, here we are not going to treat them as there will be outliers in real case scenario (in Flight_Distance, DepartureDelayin_Mins and ArrivalDelayin_Mins) and we would want our model to learn the underlying pattern for such customers.

Data Preparation for Modeling

Missing-Value Treatment

• We will use KNN to impute missing values in CustomerType, TypeTravel, ArrivalDelayin_Mins, Departure.Arrival.time_convenient, Food_drink, Onboard_service.
• NImputer: Each sample's missing values are imputed by looking at the n_neighbors nearest neighbors found in the training set. Default value for n-neighnors=5
• KNN imputer replaces missing values using the average of k nearest non-missing feature values.
• Nearest points are found based on euclidean distance.

imputer = KNNImputer(n_neighbors=5)

# defining a list with names of columns that will be used for imputation
reqd_col_for_impute = [
    "Age",
    "CustomerType",
    "TypeTravel",
    "ArrivalDelayin_Mins",
    "Departure.Arrival.time_convenient",
    "Food_drink",
    "Onboard_service",
]

df[reqd_col_for_impute].head()

	Age	CustomerType	TypeTravel	ArrivalDelayin_Mins	Departure.Arrival.time_convenient	Food_drink	Onboard_service
0	65.000	Loyal Customer	Personal Travel	0.000	0.000	0.000	3.000
1	15.000	Loyal Customer	Personal Travel	0.000	0.000	0.000	NaN
2	60.000	Loyal Customer	Personal Travel	0.000	NaN	0.000	1.000
3	70.000	Loyal Customer	Personal Travel	0.000	0.000	0.000	2.000
4	30.000	Loyal Customer	NaN	0.000	0.000	0.000	5.000

# make a copy
data1 = df.copy()

# we need to pass numerical values for each categorical column for KNN imputation so we will label encode them
Gender = {"Male": 0, "Female": 1}
data1["Gender"] = data1["Gender"].map(Gender)

CustomerType = {"disloyal Customer": 0, "Loyal Customer": 1}
data1["CustomerType"] = data1["CustomerType"].map(CustomerType)

TypeTravel = {"Business travel": 0, "Personal Travel": 1}
data1["TypeTravel"] = data1["TypeTravel"].map(TypeTravel)

Class = {"Business": 0, "Eco": 1, "Eco Plus": 2}
data1["Class"] = data1["Class"].map(Class)

data1.head()

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
0	1	1	65.000	1	1	265	0	0.000	1	0	...	3	3.000	0	3	5	3	2	3.000	2.000	1.000
1	1	1	15.000	1	1	2138	0	0.000	1	0	...	2	NaN	3	4	4	4	2	4.000	2.000	1.000
2	1	1	60.000	1	1	623	0	0.000	1	0	...	1	1.000	0	1	4	1	3	2.000	2.000	1.000
3	1	1	70.000	1	1	354	0	0.000	1	0	...	2	2.000	0	2	4	2	5	3.000	4.000	1.000
4	0	1	30.000	NaN	1	1894	0	0.000	1	0	...	2	5.000	4	5	5	4	2	4.000	2.000	1.000

5 rows × 26 columns

Imputing missing values

data1[reqd_col_for_impute] = imputer.fit_transform(data1[reqd_col_for_impute])

### checking missing values
data1.isna().sum()

Gender                               0
CustomerType                         0
Age                                  0
TypeTravel                           0
Class                                0
Flight_Distance                      0
DepartureDelayin_Mins                0
ArrivalDelayin_Mins                  0
Satisfaction                         0
Seat_comfort                         0
Departure.Arrival.time_convenient    0
Food_drink                           0
Gate_location                        0
Inflightwifi_service                 0
Inflight_entertainment               0
Online_support                       0
Ease_of_Onlinebooking                0
Onboard_service                      0
Leg_room_service                     0
Baggage_handling                     0
Checkin_service                      0
Cleanliness                          0
Online_boarding                      0
Facilities_avg                       0
Online_service_avg                   0
InFlight_avg                         0
dtype: int64

• All missing values have been treated

Feature Scaling

I need to bring all of the features of a Machine Learning problem to a similar scale or range. Feature scaling can have a significant effect on a Machine Learning model’s training efficiency and can improve the time taken to train a model.

r_scaler = preprocessing.MinMaxScaler()
r_scaler.fit(data1)

modified_data = pd.DataFrame(r_scaler.transform(data1), columns=data1.columns)
modified_data.head()

	Gender	CustomerType	Age	TypeTravel	Class	Flight_Distance	DepartureDelayin_Mins	ArrivalDelayin_Mins	Satisfaction	Seat_comfort	...	Ease_of_Onlinebooking	Onboard_service	Leg_room_service	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg
0	1.000	1.000	0.714	1.000	0.500	0.031	0.000	0.000	1.000	0.000	...	0.600	0.600	0.000	0.500	1.000	0.600	0.400	0.500	0.250	0.200
1	1.000	1.000	0.000	1.000	0.500	0.303	0.000	0.000	1.000	0.000	...	0.400	0.760	0.600	0.750	0.800	0.800	0.400	0.750	0.250	0.200
2	1.000	1.000	0.643	1.000	0.500	0.083	0.000	0.000	1.000	0.000	...	0.200	0.200	0.000	0.000	0.800	0.200	0.600	0.250	0.250	0.200
3	1.000	1.000	0.786	1.000	0.500	0.044	0.000	0.000	1.000	0.000	...	0.400	0.400	0.000	0.250	0.800	0.400	1.000	0.500	0.750	0.200
4	0.000	1.000	0.214	0.400	0.500	0.267	0.000	0.000	1.000	0.000	...	0.400	1.000	0.800	1.000	1.000	0.800	0.400	0.750	0.250	0.200

5 rows × 26 columns

Feature selection thru Chi-Square method

from sklearn.feature_selection import SelectKBest, chi2

X = modified_data.loc[:, modified_data.columns != "Satisfaction"]
y = modified_data[["Satisfaction"]]
selector = SelectKBest(chi2, k=12)
selector.fit(X, y)
X_new = selector.transform(X)
print(X.columns[selector.get_support(indices=True)])

Index(['Gender', 'CustomerType', 'Class', 'Seat_comfort',
       'Inflight_entertainment', 'Online_support', 'Ease_of_Onlinebooking',
       'Onboard_service', 'Leg_room_service', 'Baggage_handling',
       'Online_boarding', 'Online_service_avg'],
      dtype='object')

• These are the TOP 12 feactures that have influence in the customers satisfaction. I will check thru Lasso

Feature selection thru Lasso

• The Features that are the least important are the coefficients that are reduced to zero

lasso_model = Lasso(alpha=0.01)
selected_columns = list(X.columns)
lasso_model.fit(X, y)
list(zip(selected_columns, lasso_model.coef_))

[('Gender', 0.11103034320005893),
 ('CustomerType', 0.18178954305679604),
 ('Age', -0.0),
 ('TypeTravel', -0.04729928071345613),
 ('Class', -0.12816598551801273),
 ('Flight_Distance', -0.0),
 ('DepartureDelayin_Mins', -0.0),
 ('ArrivalDelayin_Mins', -0.0),
 ('Seat_comfort', 0.0),
 ('Departure.Arrival.time_convenient', -0.0),
 ('Food_drink', -0.0),
 ('Gate_location', -0.0),
 ('Inflightwifi_service', 0.0),
 ('Inflight_entertainment', 0.5292057512121757),
 ('Online_support', 0.0),
 ('Ease_of_Onlinebooking', 0.18763375309558566),
 ('Onboard_service', 0.15057828409841825),
 ('Leg_room_service', 0.10751052878768497),
 ('Baggage_handling', 0.03206291440972926),
 ('Checkin_service', 0.09325684522015301),
 ('Cleanliness', 0.0),
 ('Online_boarding', 0.0),
 ('Facilities_avg', 0.0),
 ('Online_service_avg', 0.11879708455249778),
 ('InFlight_avg', 0.0)]

• From Lasso, the features the most importants are Inflight entertainment, Ease of booking online, Onboard service, Leg room service, Customer Type and Gender

Dropping the least important features for Analysis

• This is taking into account the correlation between the variables with the target (please refer to the heatmap) where the least correlation under 0.10 where Flight Distance, Departure Delay in minutes, Arrival Delay in Minutes, Arrival time convenient and Gate Location.
• The Outcome from Chi Square and Lasso.
• All of the models show similir results, so the features that shoul be dropped are the least importat one with least correlarion and lasso coefficient ~0.

data1 = data1.drop(
    [
        "DepartureDelayin_Mins",
        "ArrivalDelayin_Mins",
        "Flight_Distance",
        "Departure.Arrival.time_convenient",
        "Gate_location",
    ],
    axis=1,
)

Splitting data

X = data1.drop(["Satisfaction"], axis=1)
y = data1["Satisfaction"]

# Splitting data into training, validation and test set:
# first we split data into 2 parts, say temporary and test

X_temp, X_test, y_temp, y_test = train_test_split(
    X, y, test_size=0.2, random_state=1, stratify=y
)

# then we split the temporary set into train and validation

X_train, X_val, y_train, y_val = train_test_split(
    X_temp, y_temp, test_size=0.25, random_state=1, stratify=y_temp
)
print(X_train.shape, X_val.shape, X_test.shape)

(54549, 20) (18184, 20) (18184, 20)

print("Number of rows in train data =", X_train.shape[0])
print("Number of rows in validation data =", X_val.shape[0])
print("Number of rows in test data =", X_test.shape[0])

Number of rows in train data = 54549
Number of rows in validation data = 18184
Number of rows in test data = 18184

Encoding Categorical

X_train = pd.get_dummies(X_train, drop_first=True)
X_val = pd.get_dummies(X_val, drop_first=True)
X_test = pd.get_dummies(X_test, drop_first=True)
print(X_train.shape, X_val.shape, X_test.shape)

(54549, 21) (18184, 21) (18184, 21)

X_train.head()

	CustomerType	Age	TypeTravel	Seat_comfort	Food_drink	Inflightwifi_service	Inflight_entertainment	Online_support	Ease_of_Onlinebooking	Onboard_service	...	Baggage_handling	Checkin_service	Cleanliness	Online_boarding	Facilities_avg	Online_service_avg	InFlight_avg	Gender_0	Class_1	Class_2
34746	0.000	34.000	0.000	2	3.000	3	3	3	3	2.200	...	3	1	4	3	2.000	3.000	3.000	1	0	0
15254	1.000	23.000	0.800	3	3.000	4	1	2	5	5.000	...	5	4	5	3	4.000	3.000	3.000	0	1	0
8170	1.000	51.000	1.000	2	2.000	1	2	1	1	4.000	...	4	2	3	1	4.000	1.000	2.000	1	0	1
69801	0.800	34.000	0.000	4	4.000	1	3	3	4	4.000	...	4	3	4	5	4.000	4.000	3.000	1	0	0
58026	1.000	53.000	0.000	3	3.000	4	4	4	3	3.000	...	3	1	3	2	2.000	3.000	3.000	1	0	0

5 rows × 21 columns

Building Models

Model evaluation criterion

Model can make wrong predictions as:

1. Predicting a customer will be satisfied with the airline and the customer doesn't
2. Predicting a customer will be neutral or dissatisfied and the customer is satisfied

Which case is more important?

• Predicting that customer will be neutral or dissatisfied but he is satisfied i.e. losing on a valuable customer or asset.

How to reduce this loss i.e need to reduce False Negatives?

• Airline would want Recall to be maximized, greater the Recall higher the chances of minimizing false negatives. Hence, the focus should be on increasing Recall or minimizing the false negatives or in other words identifying the true positives(i.e. Class 1) so that the Airline can retain their valuable passangers by identifying the customers dissatisfied with the service.

First, let's create two functions to calculate different metrics and confusion matrix, so that we don't have to use the same code repeatedly for each model..

# defining a function to compute different metrics to check performance of a classification model built using sklearn
def model_performance_classification_sklearn(model, predictors, target):
    """
    Function to compute different metrics to check classification model performance

    model: classifier
    predictors: independent variables
    target: dependent variable
    """
    t0 = time.time()
    # predicting using the independent variables
    pred = model.predict(predictors)

    acc = accuracy_score(target, pred)  # to compute Accuracy
    recall = recall_score(target, pred)  # to compute Recall
    precision = precision_score(target, pred)  # to compute Precision
    f1 = f1_score(target, pred)  # to compute F1-score
    roc_auc = roc_auc_score(target, pred)
    time_taken = time.time() - t0

    # creating a dataframe of metrics
    df_perf = pd.DataFrame(
        {
            "Accuracy": acc,
            "Recall": recall,
            "Precision": precision,
            "F1": f1,
            "ROC/AUC": roc_auc,
            "Time_taken": time_taken,
        },
        index=[0],
    )

    return df_perf

def confusion_matrix_sklearn(model, predictors, target):
    """
    To plot the confusion_matrix with percentages

    model: classifier
    predictors: independent variables
    target: dependent variable
    """
    y_pred = model.predict(predictors)
    cm = confusion_matrix(target, y_pred)
    labels = np.asarray(
        [
            ["{0:0.0f}".format(item) + "\n{0:.2%}".format(item / cm.flatten().sum())]
            for item in cm.flatten()
        ]
    ).reshape(2, 2)

    plt.figure(figsize=(6, 4))
    sns.heatmap(cm, annot=labels, fmt="")
    plt.ylabel("True label")
    plt.xlabel("Predicted label")

Model Building

models = []  # Empty list to store all the models

# Appending models into the list
models.append(("Logistic regression", LogisticRegression(random_state=1)))
models.append(("Bagging", BaggingClassifier(random_state=1)))
models.append(("Random forest", RandomForestClassifier(random_state=1)))
models.append(("GBM", GradientBoostingClassifier(random_state=1)))
models.append(("Adaboost", AdaBoostClassifier(random_state=1)))
models.append(("Xgboost", XGBClassifier(random_state=1, eval_metric="logloss")))
models.append(("dtree", DecisionTreeClassifier(random_state=1)))

results1 = []  # Empty list to store all model's CV scores
names = []  # Empty list to store name of the models


# loop through all models to get the mean cross validated score
print("\n" "Cross-Validation Performance:" "\n")

for name, model in models:
    scoring = "recall"
    kfold = StratifiedKFold(
        n_splits=5, shuffle=True, random_state=1
    )  # Setting number of splits equal to 5
    cv_result = cross_val_score(
        estimator=model, X=X_train, y=y_train, scoring=scoring, cv=kfold
    )
    results1.append(cv_result)
    names.append(name)
    print("{}: {}".format(name, cv_result.mean() * 100))

print("\n" "Validation Performance:" "\n")

for name, model in models:
    model.fit(X_train, y_train)
    scores = recall_score(y_val, model.predict(X_val))
    print("{}: {}".format(name, scores))

Cross-Validation Performance:

Logistic regression: 84.65634188822668
Bagging: 93.04662308996721
Random forest: 94.36964248837005
GBM: 92.43368287199249
Adaboost: 89.88143014972006
Xgboost: 94.6074440169235
dtree: 92.63129349345682

Validation Performance:

Logistic regression: 0.8526929260450161
Bagging: 0.930064308681672
Random forest: 0.9437299035369775
GBM: 0.9265474276527331
Adaboost: 0.8977090032154341
Xgboost: 0.9441318327974276
dtree: 0.9310691318327974

# Plotting boxplots for CV scores of all models defined above
fig = plt.figure(figsize=(10, 7))

fig.suptitle("Algorithm Comparison")
ax = fig.add_subplot(111)

plt.boxplot(results1)
ax.set_xticklabels(names)

plt.show()

Performance comparison

• Xgboost has the best performance followed by Bagging
• Bagging, Random Forest and Adaboost have outliers
• Let's tune Xgboost, Bagging and Random Forest classifier

Hyperparameter Tuning

Tuning Xgboost Ramdomized Search

%%time 

# defining model
Model = XGBClassifier(random_state=1,eval_metric='logloss')

#Parameter grid to pass in RandomSearchCV
param_grid={'n_estimators':np.arange(50,300,50),'scale_pos_weight':[0,1,2,5,10],
            'learning_rate':[0.01,0.1,0.2,0.05], 'gamma':[0,1,3,5],
            'subsample':[0.7,0.8,0.9,1]
           }
from sklearn import metrics

# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)

# To do hyperparameter tuning
from sklearn.model_selection import RandomizedSearchCV, GridSearchCV
#Calling RandomizedSearchCV
randomized_cv = RandomizedSearchCV(estimator=Model, param_distributions=param_grid, n_iter=50, n_jobs = -1, scoring=scorer, cv=5, random_state=1)

#Fitting parameters in RandomizedSearchCV
randomized_cv.fit(X_train,y_train)

print("Best parameters are {} with CV score={}:" .format(randomized_cv.best_params_,randomized_cv.best_score_))

Best parameters are {'subsample': 0.9, 'scale_pos_weight': 10, 'n_estimators': 200, 'learning_rate': 0.05, 'gamma': 5} with CV score=0.9916934529394865:
Wall time: 3min 34s

tuned_xgb1 = XGBClassifier(
    random_state=1,
    eval_metric="logloss",
    subsample=0.9,
    scale_pos_weight=10,
    n_estimators=200,
    learning_rate=0.05,
    gamma=5,
)
tuned_xgb1.fit(X_train, y_train)

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=1, eval_metric='logloss',
              gamma=5, gpu_id=-1, importance_type='gain',
              interaction_constraints='', learning_rate=0.05, max_delta_step=0,
              max_depth=6, min_child_weight=1, missing=nan,
              monotone_constraints='()', n_estimators=200, n_jobs=12,
              num_parallel_tree=1, random_state=1, reg_alpha=0, reg_lambda=1,
              scale_pos_weight=10, subsample=0.9, tree_method='exact',
              validate_parameters=1, verbosity=None)

# Checking model's performance on training set
xgb_train1 = model_performance_classification_sklearn(tuned_xgb1, X_train, y_train)
xgb_train1

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.898	0.995	0.845	0.914	0.888	0.134

# Checking model's performance on validation set
xgb_val1 = model_performance_classification_sklearn(tuned_xgb1, X_val, y_val)
xgb_val1

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.894	0.992	0.843	0.911	0.884	0.040

• Model is generalized performance

# creating confusion matrix
confusion_matrix_sklearn(tuned_xgb1, X_val, y_val)

Tuning Xgboost Grid Search

%%time 

# defining model
Model2 = XGBClassifier(random_state=1,eval_metric='logloss')

#Parameter grid to pass in GridSearchCV
param_grid2={'n_estimators':np.arange(150,250,50),'scale_pos_weight':[9,10,11],
            'learning_rate':[0.03,0.05,0.06], 'gamma':[4,5,6],
            'subsample':[0.8,0.9,1]
           }


# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)


#Calling GridSearchCV
gridsearch_cv = GridSearchCV(Model2, param_grid2, scoring=scorer, cv=5)

#Fitting parameters in GridSearchCV
gridsearch_cv.fit(X_train,y_train)

print("Best parameters are {} with CV score={}:" .format(gridsearch_cv.best_params_,gridsearch_cv.best_score_))

Best parameters are {'gamma': 5, 'learning_rate': 0.06, 'n_estimators': 150, 'scale_pos_weight': 11, 'subsample': 1} with CV score=0.9926982873125441:
Wall time: 24min 6s

tuned_xgb2 = XGBClassifier(
    random_state=1,
    eval_metric="logloss",
    subsample=1,
    scale_pos_weight=11,
    n_estimators=150,
    learning_rate=0.06,
    gamma=5,
)
tuned_xgb2.fit(X_train, y_train)

XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,
              colsample_bynode=1, colsample_bytree=1, eval_metric='logloss',
              gamma=5, gpu_id=-1, importance_type='gain',
              interaction_constraints='', learning_rate=0.06, max_delta_step=0,
              max_depth=6, min_child_weight=1, missing=nan,
              monotone_constraints='()', n_estimators=150, n_jobs=12,
              num_parallel_tree=1, random_state=1, reg_alpha=0, reg_lambda=1,
              scale_pos_weight=11, subsample=1, tree_method='exact',
              validate_parameters=1, verbosity=None)

# Checking model's performance on training set
xgb_train2 = model_performance_classification_sklearn(tuned_xgb2, X_train, y_train)
xgb_train2

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.890	0.995	0.835	0.908	0.879	0.123

# Checking model's performance on validation set
xgb_val2 = model_performance_classification_sklearn(tuned_xgb2, X_val, y_val)
xgb_val2

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.885	0.991	0.832	0.904	0.874	0.037

# creating confusion matrix
confusion_matrix_sklearn(tuned_xgb2, X_val, y_val)

• Tuned XGBoost with GridSearch is generalizing well

Tuning Bagging Ramdomized Search

%%time

# defining model
Model3 = BaggingClassifier(random_state=0, bootstrap=True)

cl1 = DecisionTreeClassifier(
    class_weight={0: 0.45, 1: 0.55}, max_depth=8, random_state=1
)
# Parameter grid to pass in RandomSearchCV
param_grid3 = {
        "base_estimator": [cl1],
        "n_estimators": [5, 7, 15, 45, 130, 145, 158, 188, 200],
        "max_features": [0.7, 0.8, 0.9, 1],
}


# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)


# Calling RandomizedSearchCV
randomized_cv3 = RandomizedSearchCV(
    estimator=Model3,
    param_distributions=param_grid3,
    n_iter=50,
    n_jobs=-1,
    scoring=scorer,
    cv=5,
    random_state=1,
)

# Fitting parameters in RandomizedSearchCV
randomized_cv3.fit(X_train, y_train)

print(
    "Best parameters are {} with CV score={}:".format(
        randomized_cv3.best_params_, randomized_cv3.best_score_
    )
)

Best parameters are {'n_estimators': 15, 'max_features': 1, 'base_estimator': DecisionTreeClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=8,
                       random_state=1)} with CV score=0.9744775512992412:
Wall time: 2min 32s

tuned_bagging3 = BaggingClassifier(
    random_state=1,
    bootstrap=True,
    base_estimator=DecisionTreeClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=8),
    n_estimators=15,
    max_features=1,
)
tuned_bagging3.fit(X_train, y_train)

BaggingClassifier(base_estimator=DecisionTreeClassifier(class_weight={0: 0.45,
                                                                      1: 0.55},
                                                        max_depth=8),
                  max_features=1, n_estimators=15, random_state=1)

# Checking model's performance on training set
bagging_train3 = model_performance_classification_sklearn(
    tuned_bagging3, X_train, y_train
)
bagging_train3

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.614	0.982	0.588	0.736	0.575	0.089

# Checking model's performance on validation set
bagging_val3 = model_performance_classification_sklearn(tuned_bagging3, X_val, y_val)
bagging_val3

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.617	0.983	0.590	0.738	0.579	0.033

# creating confusion matrix
confusion_matrix_sklearn(tuned_bagging3, X_val, y_val)

• Model is generalizing well for the recall whoever in filtering a lot of False Positives (low presicion), indicating the model is doing some mistakes predicting passangers as satisfied when in reality are dissatisfied.

Tuning Bagging Grid Search

%%time

# defining model
Model4 = BaggingClassifier(random_state=0, bootstrap=True)

cl2 = DecisionTreeClassifier(
    class_weight={0: 0.45, 1: 0.55}, max_depth=8, random_state=1
)
# Parameter grid to pass in RandomSearchCV
param_grid4 = {
        "base_estimator": [cl2],
        "n_estimators": [5,8,12,15,18,20],
        "max_features": [0.9, 1,1.2],
}


# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)


# Calling GridSearchearchCV
gridsearch_cv4 = GridSearchCV(
    estimator=Model4,
    param_grid=param_grid4,
    scoring=scorer,
    cv=5,
)

# Fitting parameters in RandomizedSearchCV
gridsearch_cv4.fit(X_train, y_train)

print(
    "Best parameters are {} with CV score={}:".format(
        gridsearch_cv4.best_params_, gridsearch_cv4.best_score_
    )
)

Best parameters are {'base_estimator': DecisionTreeClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=8,
                       random_state=1), 'max_features': 1, 'n_estimators': 15} with CV score=0.9744775512992412:
Wall time: 35.5 s

tuned_bagging4 = BaggingClassifier(
    random_state=1,
    bootstrap=True,
    base_estimator=DecisionTreeClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=8),
    n_estimators=15,
    max_features=1,
)
tuned_bagging4.fit(X_train, y_train)

BaggingClassifier(base_estimator=DecisionTreeClassifier(class_weight={0: 0.45,
                                                                      1: 0.55},
                                                        max_depth=8),
                  max_features=1, n_estimators=15, random_state=1)

# Checking model's performance on training set
bagging_train4 = model_performance_classification_sklearn(
    tuned_bagging4, X_train, y_train
)
bagging_train4

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.614	0.982	0.588	0.736	0.575	0.089

# Checking model's performance on validation set
bagging_val4 = model_performance_classification_sklearn(tuned_bagging4, X_val, y_val)
bagging_val4

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.617	0.983	0.590	0.738	0.579	0.034

# creating confusion matrix
confusion_matrix_sklearn(tuned_bagging4, X_val, y_val)

• Model is generalizing well for the recall whoever in filtering a lot of False Positives (low presicion), indicating the model is doing some mistakes predicting passangers as satisfied when in reality are dissatisfied.

Tuning Random Forest thru Randomized Search

%%time

# Choose the type of classifier. 
Model5 = RandomForestClassifier(class_weight={0:0.45,1:0.55},random_state=1)

# Grid of parameters to choose from
param_grid5 = {  
        "max_depth": [3,5,7,9,15],
        "n_estimators": [50,60,100,120],
        "max_features": [0.88,0.92,'sqrt','log2','auto'],
        "min_samples_split": [4,5,7,11,20],
        "max_samples": [0.7,0.9,None],
}


# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)
# To do hyperparameter tuning
from sklearn.model_selection import RandomizedSearchCV, GridSearchCV
# Calling RandomizedSearchCV
randomized_cv5 = RandomizedSearchCV(
    estimator=Model5,
    param_distributions=param_grid5,
    n_iter=50,
    n_jobs=-1,
    scoring=scorer,
    cv=3,
    random_state=1,
)

# Fitting parameters in RandomizedSearchCV
randomized_cv5.fit(X_train, y_train)

print(
    "Best parameters are {} with CV score={}:".format(
        randomized_cv5.best_params_, randomized_cv5.best_score_
    )
)

Best parameters are {'n_estimators': 120, 'min_samples_split': 20, 'max_samples': 0.7, 'max_features': 0.92, 'max_depth': 15} with CV score=0.9483855841371919:
Wall time: 1min 18s

tuned_rf5 = RandomForestClassifier(
    n_estimators=120,
    min_samples_split=20,
    max_samples=0.7,
    max_features=0.92,
    max_depth=15,
    class_weight={0: 0.45, 1: 0.55},
    random_state=1,
)
tuned_rf5.fit(X_train, y_train)

RandomForestClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=15,
                       max_features=0.92, max_samples=0.7, min_samples_split=20,
                       n_estimators=120, random_state=1)

# Checking model's performance on training set
rf_train5 = model_performance_classification_sklearn(tuned_rf5, X_train, y_train)
rf_train5

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.960	0.964	0.963	0.963	0.959	0.770

# Checking model's performance on validation set
rf_val5 = model_performance_classification_sklearn(tuned_rf5, X_val, y_val)
rf_val5

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.940	0.944	0.946	0.945	0.939	0.261

# creating confusion matrix
confusion_matrix_sklearn(tuned_rf5, X_val, y_val)

• Model is generalizing very good for recall and precision

Tuning Random Forest thru Grid Search

%%time

# Choose the type of classifier. 
Model6 = RandomForestClassifier(class_weight={0:0.45,1:0.55},random_state=1)

# Grid of parameters to choose from
param_grid5 = {  
        "max_depth": [12,15,18,20],
        "n_estimators": [100,120,135,140],
        "max_features": [0.88,0.92,0.95],
        "min_samples_split": [18,20,22,25],
        "max_samples": [0.6,0.7,0.8],
}


# Type of scoring used to compare parameter combinations
scorer = metrics.make_scorer(metrics.recall_score)

# Calling RandomizedSearchCV
grid_cv6= GridSearchCV(Model6, param_grid5, scoring='recall',cv=3)
# Fitting parameters in GridSearchCV
grid_cv6.fit(X_train, y_train)

print(
    "Best parameters are {} with CV score={}:".format(
        grid_cv6.best_params_, grid_cv6.best_score_
    )
)

Best parameters are {'max_depth': 20, 'max_features': 0.88, 'max_samples': 0.8, 'min_samples_split': 18, 'n_estimators': 140} with CV score=0.949825830653805:
Wall time: 4h 38min 19s

tuned_rf6 = RandomForestClassifier(
    n_estimators=140,
    min_samples_split=18,
    max_samples=0.8,
    max_features=0.88,
    max_depth=20,
    class_weight={0: 0.45, 1: 0.55},
    random_state=1,
)
tuned_rf6.fit(X_train, y_train)

RandomForestClassifier(class_weight={0: 0.45, 1: 0.55}, max_depth=20,
                       max_features=0.88, max_samples=0.8, min_samples_split=18,
                       n_estimators=140, random_state=1)

# Checking model's performance on training set
rf_train6 = model_performance_classification_sklearn(tuned_rf6, X_train, y_train)
rf_train6

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.967	0.970	0.970	0.970	0.967	0.922

# Checking model's performance on validation set
rf_val6 = model_performance_classification_sklearn(tuned_rf6, X_val, y_val)
rf_val6

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.942	0.946	0.949	0.947	0.942	0.322

# creating confusion matrix
confusion_matrix_sklearn(tuned_rf6, X_val, y_val)

• Model is generalizing very good for recall and precision but very slow learner

Model Performance Comparison

Train Performance Comparison

# training performance comparison

models_train_comp_df = pd.concat(
    [ xgb_train1.T, xgb_train2.T, bagging_train3.T, bagging_train4.T, rf_train5.T, rf_train6.T,
    ],
    axis=1,
)
models_train_comp_df.columns = [
    "Xgboost Tuned with Random Search",
    "Xgboost Tuned with Grid Search",
    "Bagging Tuned with Random Search",
    "Bagging Tuned with Grid Search",
    "Random Forest Tuned with Random Search",
    "Random Forest with Grid Search",
]
print("Training performance comparison:")
models_train_comp_df

Training performance comparison:

	Xgboost Tuned with Random Search	Xgboost Tuned with Grid Search	Bagging Tuned with Random Search	Bagging Tuned with Grid Search	Random Forest Tuned with Random Search	Random Forest with Grid Search
Accuracy	0.898	0.890	0.614	0.614	0.960	0.967
Recall	0.995	0.995	0.982	0.982	0.964	0.970
Precision	0.845	0.835	0.588	0.588	0.963	0.970
F1	0.914	0.908	0.736	0.736	0.963	0.970
ROC/AUC	0.888	0.879	0.575	0.575	0.959	0.967
Time_taken	0.134	0.123	0.089	0.089	0.742	0.922

Validation Performance Comparison

# training performance comparison

models_val_comp_df = pd.concat(
    [
        xgb_val1.T,
        xgb_val2.T,
        bagging_val3.T,
        bagging_val4.T,
        rf_val5.T,
        rf_val6.T,
    ],
    axis=1,
)
models_val_comp_df.columns = [
    "Xgboost Tuned with Random Search",
    "Xgboost Tuned with Grid Search",
    "Bagging Tuned with Random Search",
    "Bagging Tuned with Grid Search",
    "Random Forest Tuned with Random Search",
    "Random Forest with Grid Search",
]
print("Validation performance comparison:")
models_val_comp_df

Validation performance comparison:

	Xgboost Tuned with Random Search	Xgboost Tuned with Grid Search	Bagging Tuned with Random Search	Bagging Tuned with Grid Search	Random Forest Tuned with Random Search	Random Forest with Grid Search
Accuracy	0.894	0.885	0.617	0.617	0.940	0.942
Recall	0.992	0.991	0.983	0.983	0.944	0.946
Precision	0.843	0.832	0.590	0.590	0.946	0.949
F1	0.911	0.904	0.738	0.738	0.945	0.947
ROC/AUC	0.884	0.874	0.579	0.579	0.939	0.942
Time_taken	0.040	0.037	0.033	0.034	0.258	0.322

• The XGBoost, Bagging and Random Forest tuned, performed practically equal, regardless the cross-validation technique.
• The XGBoost, Bagging and Random Forest tuned, provided a great recall. Meaning the models are predicting good for satisfaction.
• Bagging is filtering out a lot of false positives, meaning the the model is predicting customers as satisfied when they are dissatisfied.
• The best model for all metrics is both Random Forest, however I would go for RF Randomized Search because it takes lesser time than the other.

Test for Random Forest Randomized

# Calculating different metrics on the test set
rf_test5 = model_performance_classification_sklearn(tuned_rf5, X_test, y_test)
print("Test performance:")
rf_test5

Test performance:

	Accuracy	Recall	Precision	F1	ROC/AUC	Time_taken
0	0.939	0.945	0.944	0.945	0.939	0.259

# creating confusion matrix
confusion_matrix_sklearn(tuned_rf5, X_test, y_test)

feature_names = X_train.columns
importances = tuned_rf5.feature_importances_
indices = np.argsort(importances)

plt.figure(figsize=(12, 12))
plt.title("Feature Importances")
plt.barh(range(len(indices)), importances[indices], color="violet", align="center")
plt.yticks(range(len(indices)), [feature_names[i] for i in indices])
plt.xlabel("Random Forest Randomized Search CV Relative Importance")
plt.show()

# importance of features in the tree building ( The importance of a feature is computed as the 
#(normalized) total reduction of the criterion brought by that feature. It is also known as the Gini importance )

print (pd.DataFrame(tuned_rf5.feature_importances_, columns = ["Imp"], index = X_train.columns).sort_values(by = 'Imp', ascending = False))

                         Imp
Inflight_entertainment 0.406
Seat_comfort           0.210
Online_service_avg     0.068
Ease_of_Onlinebooking  0.034
TypeTravel             0.031
CustomerType           0.026
Gender_0               0.025
Inflightwifi_service   0.023
Leg_room_service       0.020
Age                    0.020
Cleanliness            0.019
Food_drink             0.019
Online_boarding        0.019
Checkin_service        0.018
Baggage_handling       0.015
Onboard_service        0.013
Class_1                0.013
Online_support         0.010
Facilities_avg         0.007
Class_2                0.003
InFlight_avg           0.002

• Followed by the Feature Importance, The most important features are Inflight Entertainment, Seat Comfort and The Average for Online service variables.